Home

Fundamentals of Sensor Network Programming: Applications and

1. OS OS OS Networking Processor Processor Processor de zi 3 m 3 FIGURE 10 3 Nodes organized using geographically constrained group abstraction application This abstraction will be useful when we want all the agents sensing the same environment The exact relation between the nodes may be a tree or a directed acyclic graph 10 3 3 Application of GCG Abstraction Consider a task where we want to perform fusion of data collected over a specific ge ometric area We can employ the GCG group abstraction as a solution to this problem 2 1 In addition to low level exceptions there can be other types of exceptions such as communication holes arising in the geographic region Various types of protocol supports can then be used to solve this problem For example one may choose to food all nodes in the entire geographic area geocast strategy route into the region and food GEAR strategy that is geographic and energy aware routing food along the perimeter to route to a geographic location GPSR strategy greedy perimeter stateless routing or food even with existence of holes mobicast mobile multicast 10 3 4 n Hop Neighborhood Group n HNG The n HNG concept can be viewed as a notion capturing the idea of reaching all nodes that can possibly be reached within a certain time As before the relation between the nodes may be hierarchical a tree like structure The advantage of this pattern is that it does not require th
2. Li 1 1 1 1 1 Transmit Data Wireless FIGURE 4 1 Control flow in a traditional b event driven and c wireless platforms above a given safety threshold These broadcasted messages are received and queued by neighboring sensors and forwarded to the destination As a result of hardware constraints in sensor motes we have presented the mi croframework memory usage in Table 4 1 Table 4 1 shows the memory footprint of the components various settings for the configuration macros The major difference between a fully functional node and a standalone node is that the latter does not have a kernel and acts like a data sensing unit whereas the fully functional node has a kernel and can route and receive messages efficiently Therefore a fully functional node has more RAM and ROM space TABLE 4 1 Resource Requirements For Target Property MicaZ Mica2 Mica2Dot Flash memory kB 128 128 128 Measurement memory kB 512 512 512 EEPROM kB 4 4 4 A D channels 10 bits 8 10 bits 8 10 bits 8 Frequency MHz 1400 2483 5 433 868 916 433 868 916 Data rate kbps 250 19 2 19 2 Outdoor range m 100 300 300 Size 6x3xlcm 6 x 3 x 1 cm 2 5 x 0 6 cm www finebook ir INTRODUCTION TO SENSOR COMPUTING 43 The application of all of these concepts of sensing computing and communications are folded into a new sensor based programming paradigm which is the topic of the next section
3. Collision Recovery Subalgorithm Yes Not Recovered if randomDelay equals 1 Listen 4 1 randomDelay lt 4 0 endIf if Listen equals 1 amp mediumBusy equals True amp numTrials equals maxTrials Listen lt 0 Failure 4 TRUE idle lt 1 endIf if Listen equals 1 amp mediumBusy equals TRUE amp numTrials lt maxTrials Listen lt 0 numTrials Backoff 4 1 setTimer endIf www finebook ir 160 TECHNIQUES FOR PROTOCOL PROGRAMMING if Listen equals 1 amp idle equals 1 send RTS AwaitCTS 4 1 Listen equals 0 endIf if BackOff equals 1 amp timeout equals 1 BackOff 4 0 Listen 4 1 endIf if AwaitCTS equals 1 amp numTrials equals maxTrials AwaitCTS 4 0 Failure 4 TRUE idle lt 1 endIf if AwaitCTS equals 1 amp numTrials equals MaxTrials AwaitCTS 4 0 numTrials setTimer Backoff 4 1 endIf if AwaitCTS equals 1 amp got CTS equals TRUE AwaitCTS 4 0 send Data AwaitACK lt 1 endIf if AwaitACK equals 1 amp acknowledgement equals FALSE amp numTrials equals maxTrials AwaitACK 4 0 Failure 4 TRUE idle 4 1 endIf if AwaitACK equals 1 amp acknowledgement equals FALSE amp numTrials lt maxTrials numTrials setTimer BackOff 4 1 endIf www finebook ir TRANSMITTER ROLE 161 if AwaitACK equals 1 amp acknowledgement equals TRUE AwaitACK 4 0
4. FIGURE 2 3 WSN protocol stack This protocol stack is just a simple model to allow us to analyze the problem of WSN design from a new paradigm the proper relationship among the management planes and between management planes and network layers is undoubtedly more complicated Nevertheless this model is a useful design tool The energy management plane covers network wide energy conservation and is reponsible for energy saving decisions and tactics to increase network lifetime The mobility management plane is responsible for keeping track of the physical location of network nodes and responding to changes in this layout The task management plane schedules and manages tasks For more details on this topic see the text by Chakrabarty and Iyengar 2 Resource Constrained Computing Environment Small size and low cost enable sensors to be deployed in a large number in various applications Such features lead to the following resource constraints Energy The power supply of a sensor node is from batteries such as AA coin cell and lithium batteries Thus the energy supply to sensor nodes is limited Communication sensor nodes communicate over wireless links with limited bandwidth which is of the order of a few hundred kilobits per second kbps Computation Sensor nodes operate with limited processing ability and memory capacity For instance the processor speed of a Mica mote is only 4 MHz and its random access memory RAM is
5. a Can you still drop the sensors from the aircraft b What can you do when a sensor is not able to communicate with its neighbor c How does it affect the density of deployment d What can you say about the topology of the network e How important is the node ID for a node in this situation REFERENCES 1 I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci Wireless sensor networks A survey Comput Networks 38 393 422 2002 2 K Chakrabarty and S S Iyengar Scalable Infrastructure for Distributed Sensor Networks Springer Verlag 2005 www finebook ir 26 WIRELESS SENSOR NETWORKS 3 S S Iyengar L Prasad and H Min Advances in Distributed Sensor Integration Applica tion and Theory Prentice Hall Englewood Cliffs NJ 1995 4 R Szewczyk E Osterweil J Polastre M Hamilton A Mainwaring and D Estrin Habitat monitoring with sensor networks Commun ACM 47 6 3440 June 2004 5 K Lorincz D J Malan T R F Fulford Jones A Nawoj A Clavel V Shnayder Sensor networks for emergency response Challenges and opportunities G Mainland M Welsh and S Moulton IEEE Pervasive Comput 3 4 1623 Oct Dec 2004 6 I A Essa Ubiquitous sensing for smart and aware environments IEEE Personal Commun 7 5 4749 Oct 2000 www finebook ir 3 Sensor Technology Never before in history has innovation offered promise of so much to so many in so short a time Bill Gates A typica
6. 1208 3331 ENSE dz V OVW sd WV VONWSO MEE oost 000z 00Sc 000 oose spunoi ejeq HD 250 www finebook ir CONCLUSION 251 TABLE 13 3 Time Scheduled by MAC During Useful Network Lifetime Power Aware MAC Tx Rx Idle QoS CSMA 196 80 19 Implemented by nonsensor network simulators B MAC 2 65 33 TinyOS without real time clock Application control 6 45 49 Our study and first adaptation of single hop MAC power aware QoS Time scheduling FARMS 7296 596 2396 Extended version tunable close multihop MAC to design constraint at idle lt 1 pAh in Fig 13 11 Therefore all necessary code porting is performed via a parser code generator and script files That the programming language is C allows for a much simpler code generator 13 4 2 Sensor Stack The sensor stack 4 can be simulated on the real node as shown in Fig 13 12 thus allowing for real data to be generated This can be used for various purposes including verification of a sensor model One caveat is that an application that requires strict timing for its data would not be accurate with this mode of simulation This mode is designed for applications that process sensor data with lax timing One such application is data gathering Most applications generate data until their buffer is filled and only then do they transmit the data This mode can also work if the application requires samples at a fairly low rate suc
7. A neglectful node will sometimes arbitrarily fail to forward data to the next node along the routing path A compromised node could even falsely acknowledge successful transmission to the sender of the dropped packet A greedy node gives higher priority to its own messages causing a hit to network performance if a high volume of network traffic passes through it One way around this is to use data redundancy or better muliple routing paths A homing attack is an attack where adversaries attempt to discover and focus on nodes that are particularly crucial or important such as local clusterheads and key managers This type of attack can be foiled if packet headers are encrypted preventing adversaries from easily finding the source or destination of a given packet A misdirection attack involves routing data along the wrong path possi bly by advertising false routing information As a result data could be routed away from a specific node or sent toward the malicious node instead of their intended destination In a smurf attack a large amount of data is mistak enly routed toward the node being attacked overloading its network connection An approach similar to egress filtering can be used to defend against this type of attack A black hole is a node that falsely advertises zero cost routes to all other nodes causing all nearby nodes to forward packets to it and to incorrectly update their own route costs to be erroneously l
8. PowerupC Leds www finebook ir 116 SENSOR PROGRAMMING The code below shows how TinyOS has implemented the interface Leds Courtesy of TinyOS authors Phil Buonadonna David Gay Philip Levis and Joe Polastre description TinyOS Configuration File LedsC configuration LedsC provides interface Leds j implementation components LedsP PlatformLedsC Leds LedsP LedsP Init lt PlatformLedsC Init LedsP LedO gt PlatformLedsC Led0 LedsP Led1 gt PlatformLedsC Led1 LedsP Led2 gt PlatformLedsC Led2 We can now draw the component architecture for PowerupAppC which contains three subcomponents two of which were provided by the system facilitating our task of building our component PowerupAppC Note that the configuration LedsC provides an implementation for the interface Leds by further wiring Leds to LedsP using the statement Leds LedsP and the interfaces of LedsP to the interfaces of PlatformLedsC which is defined elsewhere in the system If we really want to understand how the interface Leds is actully implemented we need to explore the system further for explanations of PlatformLedsC which we will not do In order to run our program all we need is an implementation of the interface Leds and we have found one namely the configuration LedsC which provides interface Leds We can now run the program PowerupC As we discussed in the previous example when a program is run the event
9. Sensing applications are defined with unique data sampling needs To achieve this in a power aware manner a distributed WSN algorithm at the network layer is used to manage data routing from individual nodes to a central coordinator 12 1 7 Data Routing This category of routing does not have any application control at higher levels As soon as the data are sampled they are forwarded to the nearest forward node for delivery to the destination The performance of such an application is based on more efficiently forwarding the data toward the destination void main void Sensing device with reduced functions UINT32 my timer UINT8 failures ping cnt Hallnit evblnit will have to be called by the user hallnit evblnit www finebook ir MOTIVATION AND BACKGROUND apllnit init the stack conPrintConfig ENABLE GLOBAL INTERRUPT enable interrupts EVB LED1 OFF EVB LED2 OFF debug level 10 ifdef LRWPAN_ROUTER routerState ROUTER_STATE_JOIN_NETWORK while 1 apsFSM switch routerState case ROUTER_STATE_JOIN_NETWORK aplJoinNetwork routerState ROUTER_STATE_JOIN_WAIT break case ROUTER_STATE_JOIN_WAIT if apsBusy break if aplGetStatus LRWPAN_STATUS_SUCCESS conPrintROMString Network DATES ERSTER TERIS ore Nene ere gee ore Join _succeeded n printJoin Info my_timer halGetMACTimer routerState ROUTER STATE NORMAL ping_cnt 0 fa
10. v represents the graph induced by V v This is illustrated in Fig 4 36 Let G V E be a graph with k components a V If G a has more than k components a is called an articulation point In a connected graph a vertex whose removal disconnects the graph is also called an articulation point In the graph shown in Fig 4 36a vertices 4 and 7 are the two articulation points A connected graph that has no articulation point is called a biconnected graph Any cycle is biconnected Let G V E be a graph and u v be nonadjacent vertices of the same component of G A subset S of V is called a u v separator if the removal of S from G results in u and v lying on different components In other words S is a u v separator if there is no path from u to v without passing through the vertices of S A u v separator that does not contain any other u v separator is called a minimal u v separator A u v separator having a minimum number of vertices is called a minimum u v separator Let S1 1 2 3 4 S2 fl 2 4 S3 2 4 s4 2 4 8 S5 8 86 5 www finebook ir CONNECTIVITY 77 b FIGURE 4 35 a The BFS spanning tree b the DFS spanning tree Each term S j 5 2 53 4 S5 S6 is a 7 6 separator in the graph G of Fig 4 37 However 5 is a minimal 7 6 separator Similarly 5 and Sg are also minimal 7 6 separators 5 8 and Sg 5 are two minimum 7 6 separators Let G V E be a conn
11. 1000 1500 2000 Number of Nodes Comparison of execution time with 100 queries increases at a faster rate for ns2 than for SensorSimulator thus allowing for large simulation setup and more scalability in SensorSimulator than in ns2 Number of queries 10 Simulation time 300 s Network dimension varies with the number of nodes Number of nodes in region 5 14 7 FINAL COMMENTS On the basis of our studies with the IEEE 802 11 MAC and the Directed Diffusion model integrated with GEAR we conclude that our simulator is at least an order of magnitude faster than ns2 and uses memory more efficiently than does ns2 The modular structure of compound modules and the ease of configuring simulation 700000 600000 500000 400000 300000 200000 100000 0 Memory Kbytes B Number of Nodes SensorSimulator ns2 FIGURE 14 16 Comparison of memory consumption www finebook ir 272 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION scenarios via an initialization file offers us a tremendous amount of flexibility to model and study the dynamic behaviors of both the sensor network and the application environment in which such networks are expected to operate APPENDIX This section provides practical guidelines for SensorSimulator software the source code for which can be found at http csc lsu edu sensor web What is SensorSimulator As discussed in the main text of this
12. 8 1 How does the S MAC differ from traditional wireless MAC 8 2 What are the modes of operation of S MAC www finebook ir 154 ALGORITHMS FOR WIRELESS SENSOR NETWORKS 8 3 Whatis idle listening 8 4 What are the sources of energy drain in a sensor node 8 5 What is synchronization Show an implementation in which five nodes syn chronize with each other 8 6 Describe the major contents of a routing table and give an example of nodes in a network slowly building their routing neighborhood routing table 8 7 Provide an example where data are routed using the previous routing table and describe how a routing table changes over time 8 8 Describe the CSMA CA mechanism 8 9 Describe the TDMA mechanism and its advantages and disadvantages for sensor network applications REFERENCES 1 Crossbow Reference Manual Courtesy Crossbow Inc 10 11 J Agre L Clare and S Sastry A taxonomy for distributed real time control systems Adv Comput 49 303 352 1999 K Chakrabarty and S S Iyengar Scalable Infrastructure for Distributed Sensor Net works Springer Verlag 2005 J Hill R Szewezyk A Woo S Hollar D Culler and K Pister System architecture directions for networked sensors in In Architectural Support for Programming Languages and Operating Systems 2000 pp 93 104 S S Iyengar R L Kayshyap and R N Madan Distributed sensor networks IEEE Trans Syst Man and Cyb
13. A Demers S Shenker and L Zhang Macaw A media access protocol for wireless LANS Proc ACM SIGCOMM 1994 1994 Black G and V Vyatkin Intelligent component based automation of baggage handling systems with IEC 61499 JEEE Trans Autom Sci Eng 6 2009 Bondy J A and U S R Murty Graph Theory with Applications North Holland NewYork 1976 Brooks R R and S S Iyengar Multi Sensor Fusion Prentice Hall Englewood Cliff NJ 1997 Cassandras C G and S Lafortune Introduction to Discrete Event Systems Kluwer Academic Jan 1999 Chakrabarty K and S S Iyengar Scalable Infrastructure for Distributed Sensor Networks Springer Verlag 2005 Chan H A Perrig and D Song Random key predistribution schemes for sensor networks Proc IEEE Security and Privacy Symp 2003 May 2003 Chandrasekharan N and S Iyengar NC algorithms for recognizing chordal graphs and k trees IEEE Trans Comput 37 10 1988 Crossbow Imote2 Datasheet courtesy Crossbow Technologies Crossbow MIB520 Datasheet courtesy Crossbow Technologies Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 307 www finebook ir 308 BIBLIOGRAPHY Crossbow Moteworks Software Reference Manual courtesy Crossbow Technologies Crossbow Product Feature Reference Manual courtesy Crossbow Technologies
14. Hofmeyr S A S Forrest and A Somayaji Intrusion detection using sequences of system calls J Comput Security 6 3 151 180 1988 Hopcraft J E R Motwani and J D Ullman Introduction to Automata Theory Languages and Computation 2nd ed Addison Wesley Nov 2001 http blog xbow com xblog sensorboards http inst eecs berkeley edu cs 194 5 sp08 lab 1 index html http www cs rpi edu cheng3 sense http www isi edu nsnam ns ns documentation html IEEE Standard Dictionary of Electrical and ElectronicTerms 6th ed IEEE 1997 Ilgun K R A Kemmerer and P A Porras State transition analysis A rule based intrusion detection approach IEEE Trans Software Eng 21 3 151 180 March 1995 Intanagonwiwat C R Govindan D Estrin J Heidemann and F Silva Directed diffusion for wireless sensor networking IEEE ACM Trans Networking 11 1 216 Feb 2003 Iyengar S S and R R Brooks eds Distributed Sensor Networks 2nd ed CRC Press Dec 2004 Iyengar S S and R Brooks eds Distributed Sensor Networks CRC Press 1995 Iyengar S S R L Kayshyap and R N Madan Distributed sensor networks EEE Trans Syst Man Cyber 21 5 1027 1031 1991 Iyengar S S L Prasad and H Min Advances in Distributed Sensor Integration Application and Theory Prentice Hall 1995 www finebook ir BIBLIOGRAPHY 309 Iyer V R M Garimella Rama Murthy and M B Srinivas Min loading ma
15. More recent advances in micromechatronics systems and microfabrication tech nology have led to the availability of low cost low power multifunctional modern sensors A modern sensor consists of a sensing device s micro controller onboard memory and a transceiver The structure of a sensor node is dependent on the ap plication In general a sensor node consists of four basic components sensor unit processing unit transceiver unit and power unit A sensing unit is usually composed of sensors and analog to digital converters ADCs The analog signals are produced Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 15 www finebook ir 16 WIRELESS SENSOR NETWORKS by the sensors that observe the phenomenon These signals are converted to digital signals by the ADC and then sent into the processing unit The processing unit in structs the sensor node to carry out the assigned sensing tasks and manages the sensor node s collaboration with the other nodes A processing unit has enough storage for the real time operating system protocols and other application specific algorithms The transceiver unit connects the node to the network The power unit supplies the energy for the sensor node which may be batteries coincell lithium etc or solar cells According to the design purpose sensor
16. They are equipped with physical sensors to accurately measure environmental values and transmit only when queried or when an event has triggered idling for most of its lifetime The 802 15 4 stack code is designed as a state machine that allows a reentrant API library for sending nonblocking messages The programming APIs cover each network layer individually such that it is documented as a library function these APIs are compiled into native target processors with hardware specific optimizations 11 21 Application Layer The application layer is defined by the ZigBee specification but is implemented by manufacturers It consists of three critical components Application support sublayer Applicationframework Application object ZigBee device object The application support sublayer manages communication between the various ap plication objects Each application object provides specific sensing functionality for www finebook ir 210 STANDARDS FOR BUILDING WIRELESS SENSOR NETWORK APPLICATIONS the ZigBee device The ZigBee device object defines the role of each device in the network as an end device or a network coordinator and also is responsible for the discovery of new devices and their offered services In any network there can be only one ZigBee network coordinator but multiple intermediate routers and end devices 11 2 2 Network Layer The network layer houses mesh network abstractions allowing ZigBee devices to form resi
17. allowing a programmer to manage various services of the system The two basic software components abstractions that constitute TinyOS are described below 5 1 COMPONENTS OF TinyOS TinyOS provides software abstraction for hardware components such as its commu nication routing sensing and storage subsystems Software components in TinyOS refer to abstractions of specific services either provided by either another software component or a hardware component A software component consists of any number of the following Modules Configurations 5 1 4 Modules Modules are the lowest form of abstraction provided by the TinyOS operating system They implement program logic that directly addresses a software component and can also provide a particular set of services thereby enabling the reuse of software components Other software components can interact with a module through its defined interface which specifies a set of operations services that a particular module implementation provides 3 5 1 2 Configurations On the other hand abstractions of multiple modules and other configurations grouped together form a newly abstracted component referred to as a configuration A config uration can be visualized as a supercomponent consisting of several subcomponents to provide a single unified interface Configurations wire a set of components de fined in a component signature this is the set of interfaces that a component uses and provid
18. call Leds led2Toggle In this illustration of the collection tree protocol every node in the network plays the role of either a leaf node or a root node Roots advertise themselves throughout the network while leafs create routes to these roots using a routing gradient The interface RootControl we use in Section 10 2 1 provides the necessary functions to select roots and a special implementation of the AMSend interface allows leaf nodes to forward data to these root nodes 10 3 COLLABORATION GROUP ABSTRACTIONS We can view a collection of sensor nodes as a three tuple lt S R M gt where S is the set of nodes R is a relation among the nodes and M is the set of functions that the node performs Each component of the tuple can be dynamically modified that is the set of nodes can vary as old nodes crash and new nodes are added the relationship amongst them can be modified according to new situations and new www finebook ir COLLABORATION GROUP ABSTRACTIONS 203 al a2 a3 a4 as a6 a7 FIGURE 10 2 Typical operation of a node in this hierarchy functionalities may be added to the nodes Abstract communities can be defined and abstract routing protocols may be implemented using node level protocols This also makes programming easier as the programmer at any time only needs to deal with abstract groups By having more than one group of nodes perform the same set of function
19. extract r display message m wait for receive from neighbour and send request Node P joriginator of query to node X P does not know the id of X A node P wants to know about the event e www finebook ir QUERYING IN RUMOR ROUTING Table T build jneighbourhoodtable a randomly choose ja neighbour from T packet p query q event e send j own jaddress to a packet p programming abstraction wait for reply r from node a event E extract r display message m Node Q a node that does not know anything about X so only forwards the request to a random neighbour like node R 1 flag 0 Table T build jneighbourhoodtable if flag 1 display Already forwarded else programming abstraction receive jfrom jneighbour query packet p sender P extract 4id4of sender p event E extract event p a randomly 4choose 4a neighbour from T such that the neighbour is not the sender node P programming abstraction send jrequest own jaddress to a event E wait for reply programming abstraction wait for packet p from node a sendreply to P packet p flag 1 Node C www finebook ir 192 TECHNIQUES FOR PROTOCOL PROGRAMMING a node that knows about the event node X via a neighbour I have not forwarded this message before flag 0 Table T build jneighbour
20. gt threshold call ledOToggle 9 15 QUERYING IN RUMOR ROUTING For example in Fig 9 4 node P wants to know the event that occurred at node X and sends a query The sequence of events that take place are as follows www finebook ir 190 TECHNIQUES FOR PROTOCOL PROGRAMMING 1 Node P randomly chooses Q and sends the query 2 Since Q doesn t know the answer it forwards it to R 3 Node R forwards it to C 4 Node C replies with the details of event e as well as the path to the sensor X The program below classifies the nodes into four types the querying node P the destination node X the node that has a path to the destination node C and the node that does not know about the destination node Q Note that this type of classification is implicit rather than explicit by virtue of the code that each node executes in the network Each node executes its function in the protocol and all the nodes cooperatively execute the overall protocol scheme where we have used the following programming abstractions wait for receive from neighbour and send request Node P joriginator of query to node X P does not know the id of X A node P wants to know about the event e Table T build jneighbourhoodtable a randomly jchoose j4a neighbour from T packet p query q event e sendirequest own jaddress to a packet p programming abstraction wait for reply r from node a event E
21. sink then flood network with request establishing path to sink receive if self sink amp reply received then accept it if self intermediate node on the path amp have not already forwarded the request then forward the request and record the path if self intermediate node on the path amp have not already forwarded the answer then forward the answer and store the answer We have thus far introduced several basic protocols and discussed their implementa tion details We are now ready to design some real world applications where we can use the techniques that we have learned so far and we will be discussing some of them in the following chapter 9 12 FLOODING In WSN applications sending data from one point node to all nodes in the network is one of the necessary basic behaviors This happens in situations such as event www finebook ir 182 TECHNIQUES FOR PROTOCOL PROGRAMMING monitoring where a node wants to inform some other node called the sink node in the network about the event that it has detected but has no knowledge of One solution to this problem is that node A can choose to flood the network with the data that it has at hand In flooding node A transmits its data to all its neighbors with a request that each of its neighbors should forward the data to their neighbors with a similar request to forward This continues until all the nodes in the network have received the data and have forwarde
22. therefore we must specify which components provide the particular implementation of these interfaces since any of these interfaces could be provided by multiple components Which interface is implemented by which component is spec ified by wiring 1 Each component contains a configuration file in which the wiring details are specified An example of a typical wiring is shown below PrintClientC PrintMessage PrintMessageC PrintMessage PrintClientC PrintMessage is said to have been wired to PrintMessageC PrintMessage It signifies the fact that the interface PrintMessage used in the application PrintClientC uses the implementation of this interface pro vided by the module PrintMessageC Similarly the wiring statement Print ClientC Boot MainC Boot shows that the Boot interface used by Print ClientC is provided by Boot in MainC where MainC is the system component This wiring is indicated in a module called the configuration module described below PrintClientAppC nc The configuration module first lists all the components whose names appear in this module and then gives the wiring connections as we discussed above Wiring is discussed in more detail in Section 6 2 5 With this our application code is complete and we are now ready to execute the program By convention the TinyOS system invokes a command some where which results in the invocation of the event booted implemented in the module PrintClientC Thus the syst
23. A 0 0 sensor visualization clear vis was called 1 end while datenum tmp2 tmp 4 6 lt time amp amp strcmp getdata currentline 1 A 0 0 sdisp handle num2str currentline handle sensor event currentline A struesize currentline currentline 1 tmp datevec getdata currentline 1 A end if vis_was_called pause 5 oe F getframe This line and next line meant for saving frames in file imwrite F cdata frames frame_ regexprep datestr time png 1 oo www finebook ir 288 MATLAB SIMULATION OF AIRPORT BAGGAGE HANDLING SYSTEM 2 end or pause 3 or some kind of check against current time time time 1 24 60 end percent bags missed flight 100 bags missed flights bags checked in for flights avg time to storage mean times from check in to storage 24 60 avg time in loop mean times waiting for scan 24 60 getdata m function s getdata line col file x 5 line 1 col if x gt size file 1 s 1 o 0 else s file x end handle sensor event m function handle sensor event line file global time global tmp2 global weightthreshold global bags checked in for flights global bags missed flights global times from checkin to storage global times waiting for scan event type str2num getdata line 3 file switch event type case 01 photograph person who arrived with bag disp Photo taken for owner of b
24. Crossbow Reference Manual courtesy Crossbow Technologies Crossbow Telosb Datasheet courtesy Crossbow Technologies Eckmann S T G Vigna and R A Kemmerer Statl An attack language for state based intrusion detection Proc ACM Workshop on Intrusion Detection Nov 2000 Eschenauer L and V D Gligor A key management scheme for distributed sensor networks Proc 9th ACM Conf Computer and Communication Security Nov 2002 pp 41 47 Eskin E and W Lee Modeling system calls for intrusion detection with dynamic window sizes Proc DISCEX II 2001 Fall K and Varadhan Ns 2 Network Simulator Technical Report Univ California Berkeley 2004 Fishman G S Principles of Discrete Event Simulation Wiley 1978 Forrest S C Warrender and B Pearlmutter Detecting intrusions using system calls Alter native data models Proc 1999 IEEE Symp Security and Privacy IEEE Computer Society 1999 pp 133 145 Gislason D ZigBee Resource Guide Webcom Communication Corpo 2008 Global Sensor Networks GSN Team http sourceforge net projects gsn Hill J R Szewczyk A Woo S Hollar D Culler and K Pister System architecture directions for networked sensors ACM Sigplan Notices 35 93 104 2000 Hill J R Szewczyk A Woo S Hollar D Culler and K Pister System architecture directions for networked sensors n Architectural Support for Programming Languages and Operating Systems 2000 pp 93 104
25. Datalink Reliability Data must be not only available but also accurate In wire less sensor networks a node needs to not only communicate with its neighbors also forward the periodic sensed data over the network Many of the MAC protocols are designed for efficient ad hoc communications but not for reliable data sensing as the radio does not have a way to filter floor noise or a new sensed value in harsh environments For this reason a twoway handshake is necessary between the MAC and the datalink layer which allows them to re liably capture the new data everytime a data aggregation is performed With this reliable datalink mechanism the clusterhead can further fuse the data from neighboring sensors and discard any false values 14 31 Routing In a sensor network stack the network layer is solely responsible for route planning and maintenance Most of the energy used by the network is due to its routing activity In implementing routing there are two methods one at the network layer which is controlled by distributed algorithms to form clusters and uses efficient clusterhead selection and another at the MAC layer which uses multihop routing to forward data at the lower layers by using best effort QoS 1 4 3 Data Aggregation In a large sensor network deployment many parameters are sensed over a wide area and are periodically sent to the central coordinator As the sensed parameters are the same at every node similar sensor types are attac
26. Distributed techniques consist of either separate but interdependent parts or autonomous agents that can share information IEC 61499 is a standard for distributed systems 15 2 BACKGROUND Sensing devices are central to the functioning of airport baggage handling systems BHSs Almost all airport BHSs employ the following sensors sensing devices Radar transponder Camera Smart cards Radiofrequency identification RFID Motion sensor e Automatic target recognition ATR e X ray scanner Hence the sensor network is heterogeneous The area of data fusion deals with com bining data from different kinds of sensors There are two approaches to processing Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright 2011 John Wiley amp Sons Inc 277 www finebook ir 278 MATLAB SIMULATION OF AIRPORT BAGGAGE HANDLING SYSTEM sensor 1 SSS Data sensor 2 oa Fusion Source H1 HO Center sensor n FIGURE 15 1 Flowchart showing sensors processing data signals when employing multiple sensors 1 In the first case the raw data can be transmitted to a central processor This approach requires low latency transmissions and considerably high bandwidth In the second case some or the entire signal pro cessing is performed by sensors This approach addresses latency and bandwi
27. In the future advances in microelectromechanical systems MEMSs will lead to miniature sensing devices of about 20 jum micrometers to a millimeter in length These devices will be self powered allowing even more collaboration with other devices In regard to software more standards specifying how data can be exported between different sensor networks will be established allowing a more enriched and integrated sensing experience such as Real time collaboration between navigation systems and traffic monitoring sen sors Current information about seat availability at local restaurants or physicians offices Real time environmental awareness by a wide range of applications and devices leading to better management of scarce resources such as smart energy saving homes In this regard the principles addressed in this book will serve as building blocks for developing large scale longlived systems requiring self organization and adaptivity PROBLEMS 11 Define the following a Sensor 1 2 1 3 1 4 1 5 1 6 b Ad hoc network c Distributed sensor network d Wireless sensor network e Reusability index Discuss some of the design challenges that set wireless sensor networks apart from conventional networks Crossbow Technology Inc s MTS400 multisensor board is one of the most popular multipurpose heterogeneous sensing devices available on the market Research and prepare a two page report discussing t
28. PTETZ 93 v eaepreoor vDI9TJ pue An3 a8e38eq o Aronb jo sy nsalsseq TO E PTET 99 7 CIEPTCOOT PPTOETZJ sarnser syredop ausi LO 0 0 vPI9T13 iesessn pezuogineu LAA TV AAA 39 0 ed A3 eaep 99 Zz eaeprieoo b5eq saTinsedd oued uy 90 UA I9JSUEJ UT so u A ST 9 89 10JS UI JOU o 338S 19S F0 edA4 eqeD 33 z eaepreoo b5eq sar nsedg4 d8vIO S Jo INO JUM Seg vo ppo ST 99 10 S UL o 3WIS 19S v0 edA4 eqep 33 z eaeprieoori BPeq sainseag oSe10js ojur juoA Beg 0 uorooedsut o 9 0 Seq JODIIPOY 07i p DT TJ ueos Ajunoog 0 ysop 1 uosied jjopy prousequLn v eaeprieoo UIYSIOM ZO Seq YIM poanie oym uosiod qude130jouq on sKeA y upoyo IO uonoy Aron uondiissoq 10g IBO LUIJU s jo pueg JUIAY 10 opooopnesd T SI AIAVL 285 www finebook ir 286 MATLAB SIMULATION OF AIRPORT BAGGAGE HANDLING SYSTEM i J 07 May 2009 07 14 00 mE Level amp m 011 b a Inspection mn Scanner B Induction into BHS P Sensor A rm Sorting WARNING Bag 007 loaded onto flight 752 never passed security inspection ili FIGURE 15 8 BHS flowchart showing potential security breach 15 5 SOURCE CODE The MATLAB source code for the baggage handling system is shown below This program describes a unique way to detect and respond to various combination of events and conditions Synthetic input data could be generated by hand or by another
29. Success 4 TRUE idle 4 1 endIf end 9 5 PROGRAMMING WITH LINK LAYER PROTOCOLS In this section we look at the programming challenges posed by the protocols at the link layer and we start with the ARQ automatic repeat request technique These protocols strive to send a packet more reliably using acknowledgments and resending 9 4 AUTOMATIC REPEAT REQUEST ARQ PROTOCOL The basic idea of ARQ protocols can be described as follows The transmitting node s link layer protocol accepts a data packet creates a new packet by adding to it a header and a check sum and transmits this packet to the receiver The receiver verifies the checksum and accepts the packet if the checksum verification was successful and sends a positive acknowledgment to the sender If the check sum verification is not successful a negative acknowledgment is sent The transmitter on receiving the positive acknowledgment will know that the message was received successfully If on the other hand the sender received a negative acknowledgment it resends the packet 9 5 TRANSMITTER ROLE Note that the sender times out and exits to go to sleep if the negative acknowl edgment is not received within a reasonable time Furthermore the sender keeps sending the message indefinitely as long as the receiver keeps sending it negative acknowledgments Let p be a data packet coming from the MAC layer frame Header data packet p t check sequence FCS Const
30. The interest message is then forwarded to the node that has a lower estimated cost to the region as calculated by the GEAR protocol 24 The next node follows the same procedure and forwards the message toward the region by geographic routing If a node in the path does not have any neighbors or all its neighbors are away from the region then it sends a message to its parent node that it is a deadend The parent node on updating the cost of the unreachable node forwards the query see Fig 14 5 for query structure in an alternate route toward the region In the target region the interest is disseminated by using recursive flooding The interest cache is maintained at each node in the path with its gradient of interest to each neighbor The nodes in the region that have the specified properties of the interest send out data Nodes that send data out are referred to as publishers The data are marked as exploratory to reinforce the path that was taken by the interest On receipt of the data marked as exploratory by the subscriber a positive reinforcement message is sent out by the subscriber node Each node on the path forwards this message thus reinforcing the path to the region When a node reinforces a path its cost to the region is known and this cost is sent back to its source node which updates the cost information of that node to the particular region of interest Thus the path with the lowest cost is always maintained reinforcing the rout
31. aggregated on the servers to archive and interface to standard IP based networks or online queryable geographic information systems GISs Wireless ad hoc sensor networks WASNs have attracted much attention in recent years from a diverse set of research communities Researchers have been concerned mostly with exploring applications scenarios investing new routing and access control protocols proposing new energy saving algorithmic techniques and developing hardware prototypes of sensor nodes PROBLEMS 11 1 Contrast the features of ZigBee standard with Bluetooth wireless protocol 11 2 In three paragraphs about 600 words describe the application layer in ZigBee stack architecture www finebook ir 11 3 11 4 11 5 11 6 11 7 11 8 11 10 11 11 11 12 PROBLEMS 213 What methods does ZigBee utilize for device discovery Describe the network layer in ZigBee stack architecture Describe the topologies supported by ZigBee Write short notes on the following ZigBee devices and components a ZigBee network coordinator b ZigBee router c ZigBee end devices In about 500 words explain the ZigBee profile Write short notes about 600 words on the ZigBee cluster library ZigBee binding and ZigBee binding table Briefly describe the lighting profile in the ZigBee home control environ ment Research and prepare short notes on switch profile in the ZigBee home automation environment Figure 11 4 shows a
32. and J D Tygar Spins Security protocols for sensor networks Wireless Networks 8 5 521 534 2002 www finebook ir 17 Closing Comments A successful application of sensor networks is to prevent possible future disasters by analyzing the sensor based data collected over long periods of time A case in point is tsunami warnings Tsunamis can be described as very long wavelength waves of water caused by a sudden displacement of the ocean bed The rate at which a wave loses energy is inversely related to its wavelength and its velocity and directly proportional to water depth Most tsunamis are caused by undersea earthquakes volcanic eruptions landslides or meteor impacts and are usually preceded by seismic disturbances In case of inland seismic events there exists a worldwide network of sensors A similar network is conspicuously absent for events originating at sea Sensor based tsunami warning systems have proved to be effective in Japan and the United States The currently operational system for tsunami detection called deep ocean assessment and reporting of tsunami DART is very useful for the tracking of tsunami warnings The distributed sensor network will certainly advance the state of the art in wireless tsunami based sensor networks by designing in expensive expendable and massively deployable innovative sensors It would be very interesting to integrate these types of sensor networks with social networks where cellular s
33. aplGetStatus conPrintROMString Waiting T PEE then trying again n my_timer halGetMACTimer wait for 2 seconds while halMACTimerNowDelta my timer MSECS TOMACTICKS 2 1000 rfdState RFD STATE JOIN NETWORK break case RFD STATE NORMAL send to some target in the tree dstADDR saddr PING_DST_SADDR send a message then sleep increment ping counter ping_cnt count every so often send an APS ack request to ensure that we are still actually associated with our parent and that the packet actually reached the coordinator APS acks require more waiting time and overhead so only use them when necessary A MAC ack is always requested for a data packet but this only ensures that the packet was received by the radio of our parent ie we have the correct short address panid of the parent if the parent has dropped us from its neighbor table for some reason then the packet is rejected at at the nwk level www finebook ir 232 INSPIRE case Also if we are going through a router s to the coordinator then the MAC ack is only good for the first hop to our parent If count conPrintROMString Requesting nes feet cS its Oh testcase eas ETE APS wack n aps ack TRUE count 0 else aps ack FALSE payload 0 BYTE ping cnt payload 1 BYTE ping_cnt gt gt 8
34. code tested in a laboratory environment to explain the concepts Most of the code is written and tested by N Paramesh and our student Chuka Okoye and some examples have been adopted and modified from nesC and TinyOS manuals The programs were tested on various network configurations consisting of six TelosB sensor motes and a laptop Currently there exist several platforms designed to run programs written for TinyOS such as Mica Telos and Intel X scale family of motes Each of these motes has a unique functionality that differentiates it from others While several of the programs that have been written to illustrate concepts in this book are platform independent most have been tested primarily on TelosB a Telos family suite of sensors TelosB motes are IEEE 802 15 4 compliant devices having an integrated onboard antenna and a range of sensing devices light temperature and humidity sensors They are characterized by their USB universal serial bus programmability hence programs can be written on a computer and transferred to the device for testing TelosB sensors have an already implemented 802 15 4 physical layer and a software MAC media access control layer that ensures that sensors can communicate with each other using the onboard antennas and existing radio drivers Programs written in the TinyOS environment can be checked for program correctness either by running them in the TinyOS simulator TOSSIM or by using the integrated LEDs ligh
35. m 18 c 3 and Rank n c 15 3 12 4 3 Nullity m n c 18 15 3 6 4 4 FIGURE 4 30 a Cities and their distances b communication links for six cities www finebook ir 74 DATA STRUCTURES FOR SENSOR COMPUTING 10 13 8 6 12 FIGURE 4 31 A disconnected graph G Notice that rank nullity number of edges Also observe that for a connected graph n 1 is the rank and m n 1 is the nullity Moreover let us consider a connected graph along with a spanning tree T with e being a chord If we include e in T it will create a cycle Here b b b are branches and e alone is the chord This cycle is called a fundamental cycle There are m n 1 fundamental cycles It is also interesting to count the number of spanning trees of a graph Let G be a graph and T be its spanning tree Let e be a chord and let e bi b2 bj be its fundamental cycle By including e with T and removing from T any b we obtain a different spanning tree denoted as 7 e For every chord e we can find these spanning trees 7 e i 1 2 We observe that any spanning tree can be obtained by repeating this operation on the given spanning tree T A graph G and all its spanning trees are given in Figs 4 33a 4 33c The operation including a chord to the spanning tree and deleting a branch from its fundamental circuit is called cycle interchange It is noted that given any two spanning trees T
36. makes it challenging to adopt authentication schemes developed for more conventional networks 16 2 4 Conclusion For a broader treatment of these subjects refer to the article by Kalindi et al 3 16 3 DENIAL OF SERVICE ATTACKS IN MULTIPLE LAYERS Sensor networks could be subjected to several kinds of attacks not limited to denial of service attacks node takeovers routing attacks and possibly attacks on the physical security of nodes This section provides an overview of potential denial of service attacks that can take place at different layers of the sensor network It is important to note that in each layer of the sensor network architecture the denial of service attack is of a different nature see Fig 16 1 16 3 1 Physical Layer Two types of denial of service attacks take place at the physical layer jamming and tampering Jamming Jamming refers to the transmission of radio signals that interfere with the communication operations of the nodes in a sensor network Jamming is a simple and effective method of disrupting communications Generating a significant amount of noise on a narrow band of frequences will disrupt the communications that use those www finebook ir DENIAL OF SERVICE ATTACKS IN MULTIPLE LAYERS 299 Task Management Plane Mobility Management Plane Energy Management Plane Application Layer Taking advantage of programming flaws by hacking the sytem Transport Layer Flooding Desynchronizing Ne
37. on going problems which may need reconfiguration through the use of low level sensor programming The coordinator node synchronizes all the member nodes and provides them the flexibility to communicate data in either in their assigned time slots or by re calculating a different slot Moreover the coordinator node acts as a router to route all data from its domain to coordinators in other domain peer to peer structure or to a higher level coordinator if it has a hierarchical topology The router nodes act like clusterhead and allow all their children to directly interface to enable routing The main function of these router nodes is to provide reliable multihop paths in the event of a failure on the network www finebook ir 212 STANDARDS FOR BUILDING WIRELESS SENSOR NETWORK APPLICATIONS The end devices are mostly sensors connected to physical devices and have very little memory requirements They are mainly designed to be compatible with different manufacturers specifications This makes it possible to mix and match different sensors without any cross compatibility issues The ZigBee Alliance provides a number of profiles that provide a framework as shown in Fig 11 2 in which related applications can work simultaneously In this way end devices from different vendors can interoperate as long as they adhere to the given profile One of these profiles is the home control lighting profile This profile focuses on sensing and controll
38. transmit frame 1 RECEIVER fired NULL receive if checksum is valid then send ve ack else send ve ack If the negative acknowledgment was lost and the sender had timeout then the receiver will end up waiting forever executing the receive packet statement See code for receiver node j above Excessive waiting results in battery power wastage 9 6 ALTERNATING BIT BASED ARQ PROTOCOLS A slight variation of the protocol above is the alternating bit based protocol which attempts to transmit a packet to its neighbor In this protocol the transmitter uses a www finebook ir 164 TECHNIQUES FOR PROTOCOL PROGRAMMING bit called the control bit which is set to 0 and 1 alternately To start with in order to transmit a packet pi the transmitter transmits prepends a 0 to Pi producing a new packet 0 p and then transmits it to the receiver sets it s timer and waits for an acknowledgment ACK If timeout occurs before receipt of an ACK the transmitter resends 0 p so that the receiver knows that the same data are being retransmitted When the ACK is received from the receiver the transmitter transmits the next packet p2 prepending it with a control bit as 1 p2 If this is acknowledged the transmitter transmits 0 p3 and so on Thus this protocol manages to transmit the data even when the ACK is lost i Transmitter i Control bit b 0 REPEAT repeat for ever read packet to send p RESEND transm
39. 100 MB MicroFrame Work 2 100B 802 14 4 Network Stack 30 50 EVENT FRAMEWORK CLUSTER BEAD SENSORS IN SENSING COORDINATOR H H 1 Li 1 H Li 1 1 1 i H H Tes i 1 n CH POLL 1 1 i FUSION INTERVAL 1 1 1 1 H 1 i i i 1 H 1 DATA lt H 1 1 c T 1 1 H i DATA H 1 H i 2 1 1 i 1 ri I 1 lt gt H Ko 1 DATA I 1 ME H I N kd E i i ya car i i ld H i i SUPRISE AGGREGATED DAT ACT N nS in i i H cs m gt K ALARM i dz 1 1 1 ss iO m css af I nh EXPECTED EVENT 1 m f EVENT C1i B H I H ha I H C3 8 K SYNC EVENT 1 1 1 i C1 c H H i H 1 1 LH 1 FIGURE 12 2 Active object frame work for sensor communications www finebook ir 220 INSPIRE 1MB 100KB 10KB RAM data 1KB 100B 10B 1KB 2KB 10KB 100KB 1MB 10MB ROM code FIGURE 12 3 Total RAM and ROM size required for microframework small RTOS and other popular RTOSs and OSs 12 1 5 MCU II Framework Specifications In a sensor network stack see Table 12 3 the network layer is solely responsible for route planning and maintenance Most of the energy used by the network is due to its routing activity Two methods are used in routing implementation one at the network layer that is controlled by distributed algorithms to f
40. 4 kB www finebook ir CHARACTERISTICS OF SENSOR NETWORKS 23 Dynamic Topology Although a sensor network is usually deployed with stationary sensor nodes the network topology is prone to frequent changes This may be caused by the failure of nodes or links power running out or sometimes the mobility of nodes New nodes may be added or old nodes removed from the sensor networks As a result techniques such as dynamic route changing are needed to adapt to network topology change Unpredictability Unpredictability in sensor networks refers to the uncertainty in the correctness accuracy of sensor data the reliability of communication links and the connectivity of networks Sensor networks are subject to such uncertainty from many sources The correctness of sensor data can be affected by the node sta tus and transmission situations Sensor nodes may fail because of lack of power or may be damaged by the uncontrollable events in the natural world such as fire and earthquakes Some types of sensors may need to be recalibrated after running for a certain period of time Environmental interference can also cause unpredictable readings from sensors The nature of sensor signal propagation and the environment for the signal propagation such as the surface roughness and the presence of reacting and obstructing objects influence data communication in wireless sensor networks The wireless communication links shared by densely deployed or high traffi
41. Applications in Sensor Deployment In graph theory two graphs G and H are said to be isomorphic to each other if there is a bijection between the vertex sets of G and H In other words although G and H may look different when drawn they are essentially the same graph One of these graphs say H can be redrawn in such a way that it looks identical to G Inpractice it is very difficult to identify isomorphism among given graphs Many ad hoc deployments of sensor networks need localization techniques to discover and form a network Several algorithms exist for efficient topology control and localization techniques for known types and structures of graphs To identify or to restructure the deployment toplogy to a known structure graph isomorphism will be useful In nesC programming knowledge of fundamental data structures that preserves certain contexts also becomes necessary as real world applications get written For example a queue or linked list can be implemented to store notifications or events in the order in which they occur The importance of data structures cannot be over emphasized since efficiency and effective usage of resources are synonymous with sensor network applications The use of inappropriate data structures can only lead to the rapid demise of existing sensor deployments and for this reason the next several sections will delve into the details of graphs www finebook ir GRAPH AND TREES 61 FIGURE 4 14 A simple ill
42. COORDINATOR aplFormNetwork while apsBusy apsFSM wait for finish conPrintROMString Nwk formed waiting for join and _reception n while 1 apsFSM www finebook ir else do aplJoinN MOTIVATION AND BACKGROUND 225 etwork while apsBusy apsF SM wait for finish if apl con con con con bre else con GetStatus LRWPAN STATUS SUCCESS PrintROMString Network TERR Join succeeded n PrintROMString My Y ShortAddress is PrintUINT16 aplGetMyShortAddress PCRLF PrintROMString Parent LADDR PrintLADDR aplGetParentLongAddress PrintROMString Parent SADDR PrintUINT16 aplGetParentShortAddress PCRLF ak PrintROMString Network _Join TEN FAILED Waiting then trying again n my timer halGetMACTimer wait for 2 seconds whi le halMACTimerNowDelta my timer lt MSECS TO MACTICKS 2 1000 while 1 ifdef LRWPAN RFD now send p while 1 ackets packet test wait for finish while a psBusy www finebook ir 226 INSPIRE apsFSM endif ifdef LRWPAN_ROUTER router does nothing just routes DEBUG_PRINTNEIGHBORS DBG_INFO conPrintROMString Router c doing its thing n while 1 apsFSM endif endif 12 1 6 WSN Sensing Applications
43. General Structure The high level structure uses a framework that has the capacity to send and receive messages Each node has a unique ID that is used to initiate each broadcast Because it is a broadcast the destination is set as FF which allows the receiving node to rebroadcast the message untill all the nodes have received the messages from all other nodes This creates numerous duplicates and chains in the network path The type of data structure that is needed to store and optimize the minimal spanning tree would need to have a unique node ID a parent ID and two lists one of which keep strack of new neighbors and another one that lists members of other already existing parents www finebook ir DEFINING DATA STRUCTURE OF SPANNING TREE PROTOCOLS 87 FIGURE 4 47 Flooding from source to destination If the totals of both the lists are counted the sum will be equal to the number of edges in the network discovered If all the edges equal maximum edges then the program can stop 4 15 3 Programming Problem Formulating a flooding based protocol for tree construction can be a basis to prove that a general structure of message passing protocol in sensor networks can be used to discover a general topology for generic sensor networks Proving the above assertion for a protocol programming we would then eliminate cycles in the network and ultimately make it possible to achieve a pathway from source to destination Fig 4 47 We
44. He has been involved with research in high performance algorithms data structures sensor fusion data mining and intelligent systems Since receiving his Ph D degree in 1974 from Mississippi State University MSU USA He has directed over 40 Ph D students and 100 master s students many of whom are faculty at major universities worldwide or scientists or engineers at national laboratories or industries around the world He has published more than 380 research papers and has authored or coauthored six books and edited seven books published by John Wiley amp Sons CRC Press Prentice Hall Springer Verlag and IEEE Computer Society Press and other publishers One of his books titled Introduction to Parallel Algorithms has been translated into Chinese His research has been funded by the National Science Foundation NSF Defense Advanced Research Projects Agency DARPA Multi University Research Initiative MURI Program Office of Naval Research ONR Department of Energy Oak Ridge National Laboratory DOE ORNL Naval Re search Laboratory NRL National Aeronautics and Space Administration NASA US Army Research Office URO and various state agencies and companies He has served on the US National Science Foundation and National Institute of Health panels to review proposals in various aspects of computational science and has been involved as an the external evaluator ABET accreditation for several computer sci ence and engineering d
45. IEEE Commun Mag 40 8 102 114 Aug 2002 www finebook ir 2 Wireless Sensor Networks The Eight Fallacies of Distributed Computing Essentially everyone when they first build a distributed application makes the following eight assumptions All prove to be false in the long run and all cause big trouble and painful learning experiences The network is reliable Latency is zero Bandwidth is infinite The network is secure Topology doesn t change There is one administrator Transport cost is zero p ON Q O NN The network is homogeneous Peter Deutsch A sensor is a device that responds to a physical stimulus heat light sound pressure etc and produces a corresponding measurable electrical signal Sensor systems have been used in military industrial and medical applications for many years In military applications sensor systems are employed for tasks such as ocean surveillance and in air to air defense which detect track and identify the targets and events These defense systems use sensors such as radar passive electronic support measures ESMs infrared identification friend foe IFF sensors and electrooptic image sensors In nonmilitary areas sensor systems are widely used in applications such as robotics automated control of industrial manufacturing systems smart buildings traffic control and management monitoring organs of the human body and surveillance of natural disasters
46. Loca B F p r 1 s if 3 gt gt Em Global DB FIGURE 15 6 Detailed flowchart of BHS architecture shown in Fig 15 5 15 3 PROPOSED ARCHITECTURE After reviewing the literature we have adopted the architecture shown in Fig 15 6 The architecture shows local databases that store relevant transactions at the airport Data are transferred to and from the global database For example when a fight leaves an airport the pertinent information in the local database is transferred to the global database 15 4 SIMULATION RESULTS AND DISCUSSION Our simulation consists of a preexisting data file as shown in Table 15 1 The data file represents data collected from an array of virtual sensors The simulation engine is shown in Fig 15 7 An event handler is required to respond to these events and to firsthand visualization When an event occurs control is transferred to the event handler The event handler executes an order and a message is printed on the console Rough pseudocode for the event handler s internal logic is shown in Table 15 2 If the event needs to be visualized firsthand a procedure is invoked The visualiza tion currently displays only sensor events that occur at the current timestamp The MATLAB code for our simulation can be found in Section 15 5 A snapshot of visualization is shown in Fig 15 8 www finebook ir 284 MATLAB SIMULATION OF AIRPORT BAGGAGE HANDLING SYSTEM T
47. Sug gest some applications where mounting sensors on mobile objects will be useful Interesting scenarios are created when sensors are made much smaller in size and are programmed to become more autonomous Smart dust refers to tiny devices that are capable of limited sensing computations and commmications capabilities with short lifetime Suggest some applications where one or many bags of smart dust can be used When wireless sensors become tiny and are deployed in very large number such as in several bags of smart dust interestingly the overall behavior of such a system will in some way behave like social systems exhibiting autonomy self control limited lifetime and intracommunications Identify the management challenges in such a social system Consider an example application and propose specific management solutions appropriate for this application Traditional programming views programs as a mapping from input values to output values Suggest some characteristics of programs written for the wireless devices Hint A sensor program spends a considerable amount of time in communication and thus must be sufficiently ingenious to manage its resources such as data and power and cooperate with other sensors to achieve the overall behavior as required by the application In a computer network each node communicates with the other nodes using a set of protocols What will be the limitations of this model when applied to
48. This uses an APS ACK so that know if the message was delivered If it fails then we assume that we have lost connection and we issue a join aplSendMSG APS DSTMODE SHORT amp dstADDR 2 dst EP O cluster is ignored for direct message 1 src EP amp payload 0 2 msg length apsGenTSN aps ack rfdState RFD_STATE_WAIT_FOR_ACK break RFD_STATE_WAIT_FOR_ACK if apsBusy break if aplGetStatus LRWPAN_STATUS_SUCCESS aps ack all is well rfdState RFD STATE SLEEP J else only try a rejoin if the aps_ack failed if mac ack failed we will keep trying until the aps_ack fails we assume that we have been disconnected Try rejoining first then a join failures 0 rfdState RFD STATE REJOIN NETWORK break www finebook ir case MOTIVATION AND BACKGROUND RFD STATE SLEEP conPrintROMString Going to sleep MTn This does a disable global interrupt aplShutdown halWaitMs IO Going to sleep is platform application dependent the halSleep function is only intended for example purposes the msecs argument in halSleep may be ignored by the HAL layer as how sleep is implemented is target dependent halSleep 4000 conPrintROMString Woke up Vn aplWarmstart ENABLE_ GLOBAL_INTERRUPT rfdState RFD STATE NORMAL break ifndef LRWPAN COORDINATOR case case RFD STATE REJOIN NE
49. a practical hybrid simulation and binary search technique for determining a near optimal input throughput operating conditions 3 Our work can be extended to provide a type of simulation similar to this one The process of transporting baggage between the aircraft and the terminal that is currently employed is labor intensive and inefficient 5 The current transportation scheme is shown in Fig 15 4 The authors propose an architecture that addresses slow baggage processing speed congestion on airport airside roads inflexibility and other problems The proposed architecture is shown in Fig 15 5 The baggage truck shown in the proposed architecture facilitates temporary storage and automated transfer at the terminal and the aircraft www finebook ir osesseq poysoyo Jo SunjeomS EST Wunoll uonnjoseJ ueeJos UO Ou u89J9S UQ e 1nduBnouu Buliu ios youeas eois ud e sejnuluJ x jnduBnouui ui peueeuos seq JO e BuiueaeJ0s 13 e sDeq seulyoew pefejep jo Jequinw e a13 jo jequinwN e jndu amp nouui euiugoeul Sqa e UORne10S o Ur oeuo UJ0JJ OUI JAVI XEN e UO JE JOS O Ul 4oeuo U10JJ euin o eJ BAY e Saa uueje yeu sBeq jo 1equunw e sag Aq peueejJos speq jo JequinN e Seuiuoeul ayes SOIAIOS J9 unoo OHO e SJojunoo podies 0 eae JO ss uieg e 1exon Jo jequunN e 9 enb ul yoayo oj ejeJ euin pem enenb xey e w pem enenb Bay e enenb ui sqaq JO JequnwN e sjeBuessed jo JequinwN e eane 1 1u JoBuesse
50. above we have assigned a single role to each node Under each synchronous operation we have used a simple collection of rules that fire at appropriate instances We use a periodic clock to send the RTS and to send the data When a reply is received we set a flag that selects which send operation should be executed next when the command ired is executed The details in pseudocode is given below Node A begin Sensor boot StartTimer Timer0 10 khz FLAGS sendRTS 4 1 receiveRTS lt 0 www finebook ir THE MEDIATION DEVICE PROTOCOL 157 sendData 4 0 receiveACK lt 4 0 receiveCTS lt 4 0 if Timer Fired equals TRUE if sendRTS equals 1 send RTS MediationDevice endIf setLowPowerListeningCycle 10 receiveCTS lt 1 if sendData equals 1 send Data sendData lt 4 0 receiveACK 1 endIf endIf if Receive receivedData equals TRUE if receiveRTS equals 1 if receiveCTS equals 1 receiveRTS 4 0 SendRTS 4 0 sendData 4 1 endIf endIf if receiveACK equals 1 SendRTS 4 1 endIf endIf end Node B begin Sensor boot query MediationDevice FLAGS receiveResponse 4 1 queryResponse 4 FALSE receive receivedData if queryResponse equals TRUE timel 4 extractTime CTS PACKET wait www finebook ir 158 TECHNIQUES FOR PROTOCOL PROGRAMMING send CTS PACKET receiveResponse 4 0 receiveData 4 1 endIf if receiveData equals 1
51. acknowledgment to B We now give the pseudocode for the protocol above Node B begin Broadcast RTS to C First Acquire Medium while true Infinite loop if senseMedium not BUSY break endIf endLoop Listen for CTS from C while true packet lt listen if packet equals CTS break endif endLoop www finebook ir end Node C begin CARRIER SENSING VERSUS DECODING 151 Send Data to C while true if senseMedium not BUSY break endIf endLoop destination C unicast data destination Wait for Acknowledgement from C while true packet listen if packet equals CTS break endIf endLoop Listen for RTS from B while true packet listen if packet equals RTS break endIf endLoop Broadcast CTS to B while true if senseMedium not BUSY break endIf endLoop CTS Broadcast to B while true packet lt listen if packet equals DATA break endIf endLoop print Data received from B Broadcast ACK to B www finebook ir 152 ALGORITHMS FOR WIRELESS SENSOR NETWORKS while true if senseMedium not BUSY break endIf endLoop acknowledgeBroadcast B end Node A begin while true packet listen if packet equals RTS break endIf endLoop B is going to send data to C Must send A to sleep by setting NVA Variable duration extractDu
52. and easy to maintain using a hierarchical addressing scheme ZigBee devices are divided into three major classes 1 ZigBee network coordinator 2 ZigBee router 3 ZigBee end devices Type 1 devices are typically addressed as node 0 acting as a communication gateway and control node There is usually only one such node in a given cluster based topology and it can centrally address up to 2 6 nodes which have sufficient resources in terms of memory and data handling capability As it is the first full function device www finebook ir ZigBee DEVICES AND COMPONENTS 209 Application Framework Zigbee Device Application Application Object Object Object Application Support Sub Layer Security Service Provider Network Layer Medium Access Layer Physical Layer FIGURE 11 2 ZigBee architecture set up on the network node 0 is responsible for topology discovery and initial assignment of addresses and identifying specific nodes as intermediate routers or end nodes to initialize the sensor network Type 2 nodes are routers whose primary capability is to forward multihop communication from the data nodes to the central coordinator gateway These nodes need to have a routing stack so that they may effectively participate in routing of the measured data from their next hop neighbors Type 3 nodes are the most adaptive as these are remotely placed on a long term basis
53. as arrays queues stacks and lists which are essential to programming For network implementation an understanding of graphs is useful to appreciate routing and message passing Thus Chapter 4 explains those data structures which are essential for programming in a wireless sensor network environment Chapter 5 explains the tiny operating system TinyOS environment and is essential to understanding the subsequent chapters It presents the structure of application programming interfaces APIs built using a nesC like structure which facilitates the readability of the examples given in the rest of that chapter For the sake of completeness and continuity Chapter 5 also includes a bare minimum description of nesC programming language In Chapter 6 on nesC programming the nesC language is formally introduced and some major concepts in the language are discussed Part III discusses and presents examples of sensor network implementation Chap ter 7 provides a basic introduction to sensor programming It discusses some of the challenges encountered when programming large numbers of sensors and some inter faces provided by TinyOS to alleviate these programming challenges Chapter 8 on algorithms for wireless sensor networks is the core and the major focus of the book It gives detailed descriptions of various algorithms and their implementation in nesC In Chapter 9 on techniques for protocol programming we discuss several protocols used in most
54. as to check the amount of time required to run the simulation Figures 14 8 and 14 9 show the performance of the two simulators ns2 vs our simulation for the setup with 10 nodes and 100 nodes generating queries For these experiments a pass through simple MAC and a simple physical layer are being considered It is can be observed that the performance of both the simulators ns2 and SensorSimulator showed similar results with fewer nodes in the network As the number of nodes in 700 600 4 7 400 SensorSimulator 300 2 ns2 200 Z 100 0 Tq umi T T T T S S S S SS SS S S TS S Number of Nodes FIGURE 14 8 Execution time for 10 queries results for SensorSimulator versus ns2 www finebook ir 268 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION 16000 14000 12000 10000 8000 6000 4000 2000 SensorSimulator Hi ns2 Time sec S S SS S eS Number of Nodes FIGURE 14 9 Execution time for 100 queries results for SensorSimulator versus ns2 the network increases SensorSimulator is able to handle the traffic and the events generated in a better fashion so as to complete the simulation in a reasonable time faster than ns2 It has been observed that ns2 ran out of memory for networks above 2000 nodes It can be also observed in the figures that the execution time for the simulation run on SensorSimulator is less than that for ns2 for t
55. booted is invoked which we have given in PowerupC which calls 1ed0On to turn the red LED on Before we close the discussion on this example let us revisit the wiring concept again The wiring A X B Y means that the interface X used in component A is provided by interface Y in com ponent B The wiring B Y A X also means the same thing Sometimes we also use the operator as shown below in a configuration component C X Y www finebook ir PROGRAMMING CHALLENGES IN WIRELESS SENSOR NETWORKS 117 This means that component C provides interface X by exporting Y where Y is the component that it uses Now that we have seen some system interfaces and their implementations as provided by the system we take this opportunity to look at some more interfaces and get acquainted with them as they often may be useful in the applications that we want to develop 7 1 2 The Timer Interface One important interface that we will often need in our implementations is Timer Courtesy of TinyOS authors Cory Sharp description TinyOS Timer Interface interface Timer lt precision_tag gt command void startPeriodic uint32 t dt command void startOneShot uint32 t dt command void stop event void fired command bool isRunning command bool isOneShot command void startPeriodicAt uint32 t t0 uint32_t dt command void startOneShotAt uint32_t t0 uint32 t dt command void command uint32 t getNow command
56. build an abstract data type called binary tree where the actual tree implemen tation may be distributed between the nodes Give an interface specification for one such tree Explain how you will implement the operations 4 9 Consider a small network N of wireless sensors Draw the network as a graph Gi Let n be the node that acts as the manager node for Ni Now consider several such networks N2 N3 each represented by a graph Go Gs with their manager nodes n ns We want to manage the entire set of networks by organizing hierarchically all their managers Show a simple man agement organization of the managers ni n2 that represents a hierarchy In this hierarchy how will you incorporate the other nodes in each network Show the resulting data structure for a simple example www finebook ir REFERENCES 91 REFERENCES 1 N Chandrasekharan and S Iyengar NC algorithms for recognizing chordal graphs and k trees IEEE Trans Comput 37 10 1988 S S Iyengar and R Brooks eds Distributed Sensor Networks CRC Press 1995 P N Klein Efficient Parallel Algorithms for Planar Chordal and Interval Graphs Ph D Thesis MIT Cambridge MA 1988 4 P N Klein Efficient parallel algorithms for chordal graphs Proc 29th IEEE Symp Foun dation of Computer Section Proc 29th 1988 pp 150 161 5 S Sastry and S S Iyengar Distributed Sensor Networks CRC Press 2005 A S Tanenbaum Computer Netw
57. by a module to schedule events at a later time The structure and interface of the modules are specified using a network description language They implement the underlying behaviors of simple modules Simulation executions are easily configured via initialization files They track the events generated and ensure that messages are delivered to the right modules at the right time To take advantage of these features of OMNeT we have chosen it as the framework for sensor network simulations Its salient features include the following e OMNeT 4 4 allows the design of modular simulation models which can be combined and reused flexibly It is possible to compose models with any granular hierarchy e The object oriented approach of OMNeT allows the flexible extension of the base classes provided in the simulation kernel Model components are compiled and linked with the simulation library and with one of the user interface libraries to form an executable program One user interface library is optimized for command line and batch oriented execution while the other employs a graphical user interface GUD that can be used to trace and debug the simulation e OMNeT offers an extensive simulation library that includes support for input output statistics data collection graphical presentation of simulation data random number generators and data structures OMNeT 4 simulation kernel uses C which makes it poss
58. chapter SensorSimulator is a framework devel oped on OMNeT intended mainly to support sensor network simulations The framework provides basic modules that can be derived in order to implement the users own modules With this concept programmers can easily develop their own protocol implementations for the SensorSimulator framework without having to deal with the necessary interface and interoperability issues Overview of Framework The section gives the basic framework of SensorSimulator on OMNeT4 The user needs to know about programming in OMNeT If not you should read the OMNeT manual available at http www omnetpp org The manual then explains how to install SensorSimulator on OMNeT Installation You need a running OMNeT4 4 version 2 3 to use the LSU Sensor Simulator with all of its functionality This is available from the omnetpp org Website The user needs to go through the user manual of OMNeT to run simulations on our simulator After downloading the most recent version of the SensorSimulator from the download area of the LSU Website copy the file to the desired directory cd to this directory and then perform the following steps 1 Extract the LSU SensorSimulator tgz file It creates a LSU SensorSimulator folder 2 Append the path LSU SensorSimulator src to LD LIBRARY PATH 3 Run make in LSU SensorSimulator samples 4 You can run the simulation by executing Simulation Directory St
59. component B where c is implemented in an interface implemented in another component C See code below This is an example of fanin When C issues a corresponding event signal indicating the completion of the execution of the command c the event handling functions of both A and B are invoked and as before the order in which they are invoked is unknown An components A B C A IF gt C IF B IF gt C IF interface may be used by more than one user fan in Similarly an interface may be provided by more than one provider fan out PROBLEMS 6 1 Explain the concepts of fanin and fanout 6 2 Summarize the notion that tasks represent in TinyOS 6 3 Whatis an asynchronous command 6 4 What issues arise when atomic blocks are improperly used 6 5 Write a simple application to continually increment a counter value and send to another mote where the process is repeated 6 6 Refer to the interface folder in theTinyOS system tos interfaces 6 7 Examine the interface Leds nc and find out what each command does 6 8 Consider the interface Boot nc Why does it have only one event function 6 9 Consider the interfaces StdControl nc and SplitControl nc What difference do you find between these two interfaces 6 10 Provide a complete implementation for the data type stack and show its wiring diagram www finebook ir REFERENCE 109 6 11 Consider a simple monitoring application where we use two nodes n and n2 Each node
60. concept of developing active objects using a reusable in frastructure for specific domains such as sensor nodes and real time systems The infrastructure an example of an application framework that is referred to as mi croframework is a set of cooperating classes that makes up a reusable design The microframework captures the overall architecture for executing concurrent sensor nodes in the embedded real time environment The main element of decomposition of the microframework is an active object An active object is a state machine that executes concurrently with other objects and communicates with them by sending and receiving events This framework is typically compiled into the ROM of a senor node The nodes have reprogrammable RAM area that the application uses for running and for storage Event queues and event pools are the necessary burden you need to accept when you work with active objects The main problem with event queues and event pools is that they consume sensor nodes precious memory In order to minimize that memory usage you need to size them appropriately In this respect event queues and pools are no different from execution stacks these data structures all trade some memory for convenience of programming The correct sizing of event queues and event pools is especially important in microframework applications because the microframework offers no built in handling for over flow or under flow of events in an event pool Thes
61. convenient data structure for designing algorithms it is represented in the form of a circular array We assume that the queue is empty if and only if rear front It is easy to design procedures for insertion and deletions in a queue and so it is left to the reader as an exercise 4 5 LINKED LIST The abstract data type called a list is a sequential collection of data items called atoms along with operations to work with the collection If a1 a5 an are the atoms in a list then we write the list as a1 a2 an A list differs from a set in that 1 there is an order to the items and 2 an item may appear more than once if this is desirable in a given application www finebook ir 54 DATA STRUCTURES FOR SENSOR COMPUTING TABLE 4 3 Linked List Data Structure Name Size List 1 4 A B C D List 2 3M N O List 3 4 W X Y Z The common operations that one can perform on a list are 1 Find item check whether the item is in the list and if so indicate its position in the list 2 Insert item insert an item in the list usually in a particular location 3 Delete item delete first or possibly all occurrences of the item The flexibility in the operation definitions is intended to accommodate different environments If there is a way of comparing items aj and aj to say that a lt aj or dj gt aj or dj aj then we may define sorted lists An ascending sorted list is one where a lt a for a
62. cycles and G has n vertices and n 1 edges Any tree has at least two pendant vertices The length of a path is the number of edges that it has The length of the longest path from a vertex v to any other vertex is called the eccentricity or diameter of v and is denoted by E v E v max d v u u V 4 1 www finebook ir TREES 67 Eccentricity 5 4 343 44554 FIGURE 4 21 a Tree with radius 3 and diameter 5 b a rooted tree where d v u represents the length of the path from v to u In a tree T the vertex having minimum eccentricity is called a v to u center of the tree A tree contains one or two centers In Fig 4 21 the eccentricities of all the vertices are given Points 3 and 5 are the vertices that have the minimum eccentricity So this graph has two centers 3 and 5 The eccentricity of the center is called the radius of the tree The length of the longest path in the tree is called the diameter of the tree For the tree given in Fig 4 21a the radius is 3 and the diameter is 5 We can designate a vertex of the tree as its root If a vertex is designated as the root of the tree the tree is called a rooted tree Consider the tree given in Fig 4 21a If we designate 2 as the root of the tree the tree can be redrawn as in Fig 4 21b For a rooted tree level numbers can be defined to each of the vertices as follows The root is assigned level number 0 The vertices adjacent to the root are called the
63. data aggregation capabilities have been provided such as the collection tree protocol and dissemination protocols 10 2 1 TinyOS Data Aggregation Illustration We present the following code to illustrate TinyOS data aggregation module AdvancedSendC uses interface Boot uses interface Send as LeafSend uses interface AMSend as SerialSend uses interface StdControl as RoutingControl uses interface SplitControl as RadioControl www finebook ir DATA AGGREGATION uses interface SplitControl as SerialControl uses interface Receive uses interface Leds uses interface Read lt uint16_t gt uses interface RootControl uses interface ActiveMessageAddress as AMA uses interface Timer lt TMilli gt as Timer0 uses interface Queue lt message t gt as SerialQueue uses interface Pool lt message_t gt as SerialPool implementation Global Function Definitions 0 void serialSuccess Called whenever we send void error Called whenever an error occurs successfully void radioSuccess successfully task void sendSerialData Global Variables Declaration am_addr_t addr 1 am_group_t group 1 2000 uintl6 t sensor_value uintl6 t periodic message t buffer bool serial busy FALSE Initialize some of our other components Q0 event void Boot booted call AMA setAddress group addr if call RoutingControl start SUCCESS error if call RadioControl start SUCCE
64. during this interval Once the source node gets the CTS it will send the corresponding fragment of the network packet to the destination and wait for an acknowledgment www finebook ir EXPERIMENTAL RESULTS 265 TABLE 14 1 Parameters Used for Our 802 11 MAC Implementation Property Value SIFS 10 us DIFS 28 us Slot time 20 us Data rate 1 Mbps RTS length 44 bytes CTS length 38 bytes ACK length 38 bytes DATA length Variable Short interframe space Distributed interframe space The destination node on receiving the data frame extracts the network packet sends it to the network layer and sends back the acknowledgment to the source node Once the source node gets the acknowledgment it sends any other fragments that are to be sent to this node without any additional RTS frames Table 14 1 shows the standard parameters used for our implementation 14 6 EXPERIMENTAL RESULTS In this section we present results from three different studies First we establish that the directed diffusion simulation in our work is consistent with the diffusion implementation in ns2 Next we compare the performance with respect to execution time and memory used between our simulation and that of ns2 14 6 1 Validating Directed Diffusion Implementation In this experiment we considered sensor networks with different numbers of nodes between 5 and 200 For each sensor network we identified the maximum size of the sensor field with res
65. example the battery module needs to be informed on transmission or receipt of the packets by the physical module so that the energy consumption is updated at the node accordingly During simulation the CoOrdinator class is responsible for registering the sensor node to the sensor network Registering of the sensor node is an indication that the sensor node is up and functioning When the available energy is completely depleted the node is deregistered from the sensor network 14 4 Hardware Model 1 Battery Model This module is an essential component of the sensor node which supplies the necessary energy to the CPU module radio module and the sensors used to sense the environment Hence the battery is connected to all the hardware components of the node and its energy resource decreases depending on the power drawn by all the components At regular intervals the module updates its remaining energy depending on the type of battery model used Various models such as the linear battery model and discharge rate dependent model are being implemented When all the hardware devices report their power consumption the current discharge of the battery and hence the estimated duration T in hours which indicates how long the battery is expected to last is determined as T C I where C is the remaining capacity of the battery in ampere hours and Z is the total current drawn by the sensor node in amperes The remaining capacity in the battery can be e
66. existing sensor TABLE 3 3 Imote2 Specifications SDRAM memory 32 MB Flash memory 32 MB Frequency band 2400 2483 5 MHz Data rate 250 kbps Range of sight 30m I O ports 3 UART 2 SPI SDIO GPIO Size 36 x 48 x 9 mm Operating systems TinyOS Microsoft net Framework www finebook ir 34 SENSOR TECHNOLOGY TABLE 3 4 SHIMMER Properties Memory 10 kB RAM 48 kB ROM A D converter 12 bits Storage 2 GB microSD Communication CC2420 radio and class 2 Bluetooth Sensors MEMS accelerometer Operating life Deep sleep life gt 1 year Form factor 1 75 x 0 8 x 0 5 in network These devices may support remote access to a sensor network as in the case of Crossbow s MIB600 Ethernet interface board or basic base stations relaying data from the sensor network to computer over their USB connectivity The pur pose of gateway devices is to connect sensor nodes to existing Ethernet networks Table 3 5 summarizes the features of two popular interface boards MIB520 Fig 3 12 and MIB600 Fig 3 13 and a fully configured gateway device Stargate network 2 4 Unlike the software that runs on devices in the sensor level all programs running on the server class of devices are in an always on mode which ensures the timely translation processing and buffering of data that emanate from the wireless network These programs form the link between isolated sensing motes and the traditional Ethernet network running on an Internet client One suc
67. functions They are usually very simple devices with low resource and communication requirements It is for this reason that they can communicate only with FFDs and can never become network coordinators www finebook ir 138 ALGORITHMS FOR WIRELESS SENSOR NETWORKS t4 A n Equations to t delta propagation delay t4 ts delta propagation delay FIGURE 8 5 Receiver receiver synchronization 84 SYNCHRONIZATION IN WIRELESS SENSOR NETWORKS Time synchronization refers to a method of timekeeping requiring the coordination of multiple events to operate a system in unison It is particularly important in computing because of its critical role in communication Distributed systems rely on several synchronization mechanisms to ensure proper operation without which severe degradation in performance can occur 5 2 3 In data aggregation and event monitoring sensor networks the need for a flexible and robust time synchronization mechanism is more apparent since collaboration among nodes is critical to data reduction and the energy efficiency of sensor networks Also sensor usually monitor highly dynamic environments that change with time For this reason time is a basic requirement for nodes to correlate events that occur in the network with events in the physical world Failure to synchronize sensor nodes in a sensor network would render measurements useless since synchronization is essential for both temporal and spatial an
68. getdata line 2 file 0 0 0 event type else disp WARNING 10 Bag getdata line 2 file loaded onto _flight_ getdata line 4 file never_passed_security inspection j sensor_visualization 11 getdata line 2 file str2num getdata line 4 file NOT SCANNED event _type case 07 flight left query results field4 localdata 4 amp amp field3 01 field4 localdata 4 amp amp field3 06 gt 0 bags checked for current flight bags loaded on current flight bags_lost for x 1 line if strcmp getdata x 4 file getdata line 4 file amp amp strcmp getdata x 3 file 01 bags checked for current flight bags checked for current flight str2num getdata x 2 file end if strcmp getdata x 4 file getdata line 4 file amp amp strcmp getdata x 3 file 06 bags loaded on current flight bags loaded on current flight str2num getdata x 2 file end end bags lost intersect setxor bags checked for current flight bags loaded on current flight bags checked for current flight bags checked in for flights bags checked in for flights size bags checked for current flight 2 bags missed flights bags missed flights size bags lost 2 if size bags 1ost 2 50 disp WARNING 10 Bag s num2str bags lost not loaded onto flight 4 getdata line 4 file for x 1 size bags lost 2 sensor
69. level a sensor can be viewed as a sensing device that senses the environment measuring for example temperature pressure and smoke intensity At the higher level we view a collection of nodes as a distributed source of data streams 7 1 PROGRAMMING CHALLENGES IN WIRELESS SENSOR NETWORKS One of the main challenges in wireless sensor programming involves updating and managing several resource constrained nodes that interact in real time with the phys ical world where these nodes cannot be easily accessed We first discuss some of the system interfaces available in TinyOS 2 7 1 4 System Interfaces In the examples in earlier chapters we defined our own interfaces provided im plementation modules and wired configurations The nesC 2 language is meant primarily for programming embedded systems such as motes tiny devices in wire less sensor network applications The PrintMessage program we wrote in Chapter 6 is hardly useful in nesC applications since such programs could be written much Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 113 www finebook ir 114 SENSOR PROGRAMMING more comfortably in C So if we want to use the sensing and the wireless trans mission capability of the motes we need to get acquainted with the rich collection of built in interfaces library of
70. maintained to avoid going into a loop When a node in the region of interest receives an interest it sends back an exploratory message to the source of the interest The exploratory message follows the reverse path taken by the interest message It gets the reverse path information from the nodes When this exploratory message reaches the source node the source node reinforces the path by sending back reinforcements The reinforcements might or might not follow the same path as the initial interest message On arrival of the reinforcements the nodes in the region of interest start sending back data messages at the rate specified in the interest At regular intervals these data messages are marked as exploratory When the source receives a data message marked as exploratory it sends reinforcements to rebuild the path This would repair any holes that might have formed in the path In order to test the performance of the simulation we ran the setup with queries generated by 10 nodes at random locations in the network A similar test was per formed with 100 nodes generating queries The queries follow a multihop route to the region following the procedure mentioned above Once the query reaches the region the data are sent back once every 5 s for the complete simulation time by all the nodes in the region The objective of this kind of setup is able to check whether the simulation framework can handle the traffic generated and run to completion as well
71. more fiexible design with two microcontrollers as discussed before one dedicated to transmission and the other to data sensing As a scheduler is more dependent on sensing a real time framework is used to create specific sensing tasks that allow for the isolation of error during sensing and transmission 12 1 11 Complexity Wireless sensor systems are dedicated and unattended for most of their lifetimes so the complexity of the design needs to address the time scheduling polling syn chronization local global and communication overhead to generate the datastream real time sensor data We use specific MAC level protocols such as Berkeley MAC B MAC 4 used by TOSSIM carrier sense multiple access CSMA and the gen eral IEEE 802 11 wireless predecessor without RTS CTS These are implemented using the GlomoSIM routing layer 12 1 12 Event Driven System Wireless sensor systems are dedicated and event driven 1 2 We use a common platform that allows modules to be programmed with events A simple sensor cluster is implemented in Fig 12 2 that has five states for performing a useful task Poll is generated by the microframework which allows tracking of the current hardware clock Once the poll is active the sensor object changes states If it is idle then it will accept the poll and initialize itself with a predetermined timeout The timeout event is handled and the measured value from the sensor is read accurately www finebook
72. need to specify the protocol for nodes 0 N Initially all nodes are reset and node 0 is initialized with an ID of 0 and root r Then the parent is set to root and it broadcasts messages to all nodes 0 N Once the messages are received by other nodes on the network that are currenly reset it checks whether it has any parent ID if not it will reply ACK only to its new parent as a subscribed neighbor If a node downstream has already been assigned then it is a duplicate message and the node sends a NAK to its requesting parent All initialized parents will receive and send ACK and will add the new neighbor to their lists The protocol stops as soon as the number of neighbors is fully discovered The spanning tree is shown in Fig 4 48 4 15 4 Pseudocode Messages types M request message P Pj valid parent P is the parent of P P Pj invalid parent P has already been selected as parent node www finebook ir 88 DATA STRUCTURES FOR SENSOR COMPUTING FIGURE 4 48 Spanning tree before a nonoptimal b optimal with no chains Local data structure Neighbors set of process IDs Children Node ID Event and actions This pseudocode describes a simple way of embedding a basic concept of spanning tree data structure which was discussed in the data structures section Furthermore the structure explained in this example allows to explore a combination of simulation methodology and a code for a specific hardware
73. network at will there is no fully effec tive solution If the collisions are corrupting only a portion of the sent data packets this issue can be mitigated by using error correcting codes Some form of collision detection at the receiver could be used to detect such attacks if not to completely mitigate them An exhaustion attack is similar to a collision attack but the aim of the attack is to use up the power of network nodes and render them inoperative One way to accomplish this is to purposely cause repeated collisions and retransmissions Another way is to have a compromised node continually query other nodes for information Limiting the maximum amount of reponses that a node is allowed to make in a given time period can mitigate this problem Network delays and unreliability can be caused by the intermittent application of the abovementioned attacks In an unfairness attack malicious users transmit an unusually large number of packets if the medium is free causing other nodes to delay sending their packets This type of attack relies on abusing the cooperative priority scheme used by the MAC layer One way to mitigate this type of attack is to use small frames causing nodes to give up control of the channel after only a short time This does have overhead however and furthermore a node that cheats by vying for the channel immediately while other nodes delay randomly can still capture the channel repeatedly 16 3 3 Network Layer
74. new simulator executes at least an order of magnitude faster than ns2 while using memory more efficiently While the design we present is general the simulations focus on an implementation of the IEEE 802 11 MAC layer and directed diffusion integrated with the GEAR geographic and energy aware routing protocol The remainder of this chapter is organized as follows Section 14 3 describes the background for simulating sensor networks Section 14 4 describes the simulation problem Section 14 5 describes the implementation details of the new simulator and Section 14 6 discusses the performance of the simulator 14 3 CURRENTLY AVAILABLE SIMULATORS ns2 is a well established discrete event simulator that provides extensive support for simulating TCP IP routing and multicast protocols over wired and wireless networks 8 A radio propagation model based on two ray ground reflection approximation and a shared media model in the physical layer an IEEE 802 11 MAC protocol in the link layer and an implementation of dynamic source routing for the network layer were developed in the Monarch project 14 SensorSIM builds on ns2 and claims to include models for energy and the sensor channel 18 At each node energy consumers are said to operate in multiple modes and consume different amounts of energy in each mode The sensor channel models the dynamic interaction between the physical environment and the sensor nodes This simulator is no longer being d
75. node 2 role query generator like P fired for each randomly generated neighbour q store q s id send query to q a sendDone NIL receive u if packet received from one of the neighbours previously randomly neighbour then extract the answer halt role out side rumor domain like Q fired NIL receive if query received from a node p then Store address of p _ _for_each_randomly_generated_neighbour q __ store _q s id send query to q if packet received from one of the neighbours previously randomly chosen neighbour then extract the answer send the answer to node p role inside rumor domain like C fired NIL receive if query received from any node r then obtain answer to query from the local memory also obtain address of a node like B www finebook ir 194 TECHNIQUES FOR PROTOCOL PROGRAMMING Ei which has path to the event node like send answer address gt to node r PROBLEMS 9 1 In your own words describe the role of the mediation device protocol 9 2 Write a simple program to automatically assign random node identification numbers to sensor motes 9 3 Explain the function of the alternating bit based ARQ protocol 94 What is querying Give an example of a real world case 9 5 What are the programming challenges at the link layer Explain why 9 6 Implement a 100 random uniform distribution of nodes that are assi
76. numbers of some well known graphs are listed in the following table Sensor Number Graph Chromatic Number 1 K n 2 Tree 2 3 A graph with no edge 1 4 A cycle of even length 2 5 A cycle of odd length 3 6 Bipartite graph 2 A complete subgraph is called a clique A clique C with r vertices is called an r clique A clique C is said to be maximal if there exists no other clique that properly contains C The maximal clique of the greatest cardinality is called the maximum clique The number of vertices in the maximum clique is called the clique number In order to color an r clique we need r colors So chromatic number clique number 4 13 CLIQUE COVERING A clique cover of a graph G V E is produced by partioning V into Vj V2 Vg such that each V is a clique In Table 4 9 we show two different clique coverings of the graph G shown in Fig 4 46 The size of a clique cover V V V refers to the number of partitioned sets k The clique cover of minimum size is called the minimum clique cover A set of vertices X of a graph G V E is called an independent set if any two vertices of X are not adjacent in G Independent sets are also called stable sets A maximal www finebook ir INTERSECTION GRAPH 85 TABLE 4 9 Clique Coverings for graph G in Fig 4 46 Clique Covering 1 Clique Covering 2 Yy 1 Vi 1 2 V 10 V 3 4 5 V3 2 8 9 V3 6 7 8 V 6 7 V4 9 10 V 3 4 5 independent an
77. of Sensor Nodes 132 8 2 Distinctive Properties of Wireless Sensor Networks 134 8 3 Sensor Network Stack 135 8 4 Synchronization in Wireless Sensor Networks 138 8 5 Collision Avoidance Token Based Approach 144 8 6 Carrier Sensing Versus Decoding 148 Problems 153 References 154 Techniques for Protocol Programming 155 91 The Mediation Device Protocol 156 9 2 Contention Based Protocols 158 9 3 Programming with Link Layer Protocols 161 9 4 Automatic Repeat Request ARQ Protocol 161 9 5 Transmitter Role 161 9 6 Alternating Bit Based ARQ Protocols 163 97 Selective Repeat Selective Reject 168 9 8 Naming and Addressing 170 9 9 Distributed Assignment of Networkwide Addresses 170 9 10 Improved Algorithms 177 9 11 Content Based Addressing 179 9 12 Flooding 181 9 13 Rumor Routing 184 9 14 Tracking 188 9 15 Querying in Rumor Routing 189 Problems 194 References 194 REAL WORLD SCENARIOS Sensor Deployment Abstraction 197 10 1 Sensor Network Abstraction 197 10 2 Data Aggregation 198 10 3 Collaboration Group Abstractions 202 10 4 Programming Beyond Individual Nodes 205 Problems 205 References 206 www finebook ir x 11 12 13 14 15 CONTENTS Standards for Building Wireless Sensor Network Applications 11 1 802 XX Industry Frequency and Data Rates 207 11 2 ZigBee Devices and Components 208 11 3 ZigBee Application Development 210 11 4 Dissemination and Evaluation 212 Problems 212 Reference
78. of heterogeneity for example there may be a mix of platforms or a network may consist of a small number of spine data backbone shown to be optimal for data delivery and a large number of lower power nodes for data collection Here the backbone www finebook ir 8 INTRODUCTION nodes will have to handle different versions of their code base as well as different codebases Performance The time required to update nodes as well as tradeoffs between time and energy need to be considered Provisions to recover from faulty updates with mechanisms to verify the new software both before and during execution 14 PERFORMANCE DRIVEN NETWORK SOFTWARE PROGRAMMING There are four basic issues here 1 Quality of Service In sensor networks quality of service is an important metric to analyze the performance and reliability of different WSN routing algorithms As the sensor nodes use fixed batteries to sense and communicate it is neces sary to collaboratively use the network resources to minimize power usage and when idling conserve power by using ultra low duty cycling The communica tion module of a sensor mote uses a software stack and a radio to receive and transmit information The network stack has many layers spanning from phys ical layer to network layer with various functionalities By design a running stack needs to use a small footprint and be power aware avoiding unnecessary overheads at every layer The QoS can be de
79. of nodes randomly deployed and organized in order to achieve similar objectives see Fig 8 3 for an example These objectives are to retrieve certain data about an entity being monitored and transmit results to a remote destination serving as a data sink The random deployment of sensors and the volatile nature of the network usually due to node failure warrant the development of self configuration mechanisms to prevent network degradation and efficient transmission of information 8 2 2 Self Healing Despite measures taken to ensure durable sensor networks several factors still exist that can result in a breakdown in communication These factors include energy depletion in some key routing nodes un intentional damage by humans or other animals or even the addition of new nodes resulting in a highly dynamic network It www finebook ir SENSOR NETWORK STACK 135 Intermediate Leader Sees Data Sink Events Node g N FIGURE 8 3 Dataflow in a sensor network is for this reason that the networking stack on sensors be impervious to node additions and deletions without requiring a complete reset of the whole network 8 2 3 Dynamic Routing Communication in sensor networks can be very expensive and sensor nodes can be added and removed on the fly The need for an adaptive routing scheme based on network conditions such as link quality hop count and gradients has led to the development of several new on demand routing
80. of the kernel The scheduler can be configured to have nonpreemptive preemptive mode Non preemptive scheduling is when an intelligence surveillance reconnaissance ISR device is serviced and then the current task is run to completion In the case of pre emptive scheduling the current task is rescheduled according to the priority after the ISR has completed which might yield to another task already queued for processing 12 1 4 MCU I Framework Specifications MicroFrameWork Table 12 2 is an event driven C framework with its own multitasking kernel as shown in Fig 12 1 aimed at the mote class sensor networks Its primary advantage over TinyOS 6 is that it allows modules to be easily ported into any processor that the sensor platform uses Application level software can be easily developed using a Unified Modeling Language UML diagram and the states as shown in Fig 12 2 which represents emulation of an event model paradigm see also Fig 12 3 Input BatteryCapacity Output Residual Capacity foreachEver do Sensor ctor SensorHardware Init initialize the board MF run transfer control to MicroFrameWork end Return Data www finebook ir FARMS Application Event Processor Object FrameWork RT Kernel Target FIGURE 12 1 Stack architecture TABLE 12 2 Resource Requirements for Target Software Platforms ROM kB RAM DotNetMicroFrame Work 10 1 MB Zotta OS 15
81. of this book 41 INTRODUCTION TO SENSOR COMPUTING The availability of effective communications coupled with the computational capa bility of sensors makes it feasible to host various tasks in sensors As shown in Fig 4 2 the four primary functions are the algorithm processing process diagnostics data management and system interfaces Algorithm Processing involves tasks that are performed in a sensor node Specialized algorithms are required to condition signals encrypt data and process data in the node Depending on the overall design of the distributed sensor network DSN the nodes may implement components of a distributed algorithm The operating environment of a node is responsible for ensur ing that these algorithms are executed fairly and effectively Process diagnostics are additional computations that are performed at the sensor or cluster levels to augment the processing function of the input subsystem Various techniques for automatically embedding code in the algorithms are being investigated for diagnostics monitoring or distributed services For example such embedded code could provide status in formation and alarm data to operator monitoring stations Some diagnostic strategies require temporal information in addition to the input data Data management is another function that is becoming increasingly important for DSNs Because of the size of contemporary systems the data gathered by the collection of sensors is immen
82. pp 169 174 2 M B Srinivas V Iyer G Rama Murthy and B Hochet C error simulator for development for sensor and location aware sensing applications Proc 3rd Int Conf Sensing Technology Taichung Taiwan 2002 pp 799 804 www finebook ir 1 1 Standards for Building Wireless Sensor Network Applications Good design adds value faster than it adds cost Thomas C Gale 11 1 802 XX INDUSTRY FREQUENCY AND DATA RATES The IEEE 802 15 4 is a standard that specifies the implementation details of medium access control MAC and the physicallayer PHY for low rate wireless networks It provides the fundamental lower network layers of a wireless network focusing on low cost and low speed typically associated with wireless sensor networks Currently it has been implemented by several networking solutions such as ZigBee Wireless HART and MiWi which provide higher level communication protocols in addition to the 802 15 4 standard In this chapter most of the focus will be on the ZigBee standard due to its popularity among wireless networking vendors A simple comparison of some wireless standards is shown in Fig 11 1 The ZigBee standard defines the security networking and application frameworks for an IEEE 802 15 4 based system It creates a self forming self healing mesh network capable of supporting thousands of wireless devices on a single network 1 3 The ZigBee stack architecture is divided into five parts Security
83. priority bus scheme e g controller area network CAN can be made quasideterministic 4 5 GENERAL STRUCTURE OF PROGRAMMING This aspect has been largely ignored in the DSN literature It must cover a range of activities including designing developing debugging and maintaining programs that perform computing input and communication tasks at the sensor level Pro grams must also be developed to support distributed services that are essential for proper functioning of the DSN In addition activities such as abnormal state re covery alarming and diagnostics must be supported Figure 4 4 shows the primary functions of the programming category support for coding of the algorithm system testing diagnostics exception handling data management documentation and syn chronization A key component of each function is the differences that are imposed by having to run in a distributed environment and which services are provided by the programming language and operating system For example the algorithm at a given sensor may require data from another sensor An issue is whether the data are easily available transparent services or whether the programmer must provide code for accessing the remote data explicitly System testing diagnostics and exception handling are complicated by the fact that data are distributed and determination of the true system state is difficult Documentation includes the program source code and details of system ope
84. processData time2 4 extractTime ACK PACKET wait send ACK PACKET receiveData 4 0 endIf end 9 2 CONTENTION BASED PROTOCOLS Contention based protocols refer to a class of communication protocols that govern how multiple transmitters devices can make use of the same channel while avoiding collisions that may occur This is achieved by specifying a set of rules by which each transmitting device must abide 9 2 1 Carrier Sense Multiple Access Protocol The carrier sense multiple access CSMA protocol belongs to the class of contention based protocols in which before any information is transmitted a transmitting node verifies that no other concurrent transmissions are taking place on the shared medium The operation of a variant of the CSMA protocol with collision detection is illustrated by the state diagram in Fig 9 1 The pseudocode accompanying Fig 9 1 is provided below begin FLAGS idle 4 1 randomDelay 4 0 Listen 4 0 AwaitCTS 4 AwaitACK lt Backoff 4 0 if idle equals 1 numtrials 4 0 randomDelay lt 4 1 idle 4 0 endIf www finebook ir Transmit next bit of Frame i CONTENTION BASED PROTOCOLS 159 v Assemble a Frame v Attempt 1 Is some other station transmitting v No Transmit 1 bit of the Frame Collision Detected No v Transmission Finished es No w Y FIGURE 9 1 Carrier sense multiple access protocol Recovered
85. program they were generated by hand for testing metrics The BHS structure con sists of a main loop an event handling layer and an display layer The main loop continually scans for events and passes them to the event handler as well as being responsible for computing matrics The event handler contains the internal logic to actually generate and handle events on the basis of the input data The display layer is responsible for displaying current events visually TABLE 15 3 Performance Metrics for Test Data File Percentage bags that missed flight 11 1111 Average bag time to storage 15 3333 min Average bag time on loop 9 5000 min www finebook ir SOURCE CODE 287 snproj m clear all close all clc global time global tmp2 global bags checked in for flights global bags missed flights global times from check in to storage global times waiting for scan global weightthreshold weightthreshold 50 A textread final log txt s startdate date startdate datenum startdate time startdate currentline 1 tmp datevec getdata currentline 1 A tmp2 datevec date tmp2 tmp2 1 3 bags_checked_in_for_flights 0 bags_missed_flights 0 times_from_check_into_storage times waiting for scan while stremp getdata currentline 1 A 0 0 Swhile time lt startdate 1 disp datestr time vis was called 0 if datenum tmp2 tmp 4 6 lt time amp amp strcmp getdata currentline 1
86. read in the parameter val if the result of reading was successful i e if result SUCCESS Otherwise parameter val may contain arbitrary values It is now easy to understand the SenseC module and we now specify the wiring We need to know that Read is implemented in DemoSensorC providedbyTinyOS an instance of which has been renamed as Sensor SenseAppC This is the configuration that shows the wiring required As we did before the interface Boot is implemented by MainC Leds by LedsC Timer by TimerMilliC and Read by Sensor Courtesy of TinyOS authors Jan Hauer description Sense Application Configuration configuration SenseAppC implementation components SenseC MainC LedsC new TimerMilliC new DemoSensorC as Sensor SenseC Boot MainC SenseC Leds LedsC SenseC Timer TimerMilliC SenseC Read Sensor The StdControl interface defined in TinyOS as shown in StdControl code below is an important interface that can be used to turn on and off a component C that provides an implementation of this interface The start command can be used to turn on the component C and the stop command to turn it off Finally we have shown in Figure 7 3 the flow of control in the nesC program execution as the application executes The TinyOS system invokes the even booted which calls the command StartPeriodic The StartPeriodic command invokes the event fired which calls the command read implement
87. room layout with two doors and four lights Lights L4 Ly Ls and L4 are controlled by ZigBee enable switches S1 5 Sa and S4 respectively L1 L2 DOOR2 DOOR 1 S1 S2 S3 S4 Switches ie L3 L4 FIGURE 11 4 Problem 11 11 Design a ZigBee home automation application to perform the following tasks a Lights L and L4 should be switched ON when a person enters through door 1 b Lights L and L4 should be switched ON when a person enters through door 2 c Lights should continue to be ON when a person is present in the room d All lights should be switched OFF when a person leaves the room www finebook ir 214 STANDARDS FOR BUILDING WIRELESS SENSOR NETWORK APPLICATIONS 11 13 Discuss the equipment required e g ZigBee enabled occupancy sensors switches controllers gateways and placement of the sensors and explain in detail how you would implement the application 11 14 Describe Cyclic and Pin sleep modes in ZigBee 11 15 How do you create and maintain a list of active devices that are connected to a ZigBee network 11 16 Discuss MessageFreshener timers in ZigBee 11 17 How does ZigBee achieve low power consumption REFERENCES 1 ZigBee Alliance Overview ZigBee Alliance 2 ZigBee Wireless Networking Newnes Publications 2008 3 D Gislason ZigBee Resource Guide Webcom Communication Corp 2008 www finebook ir 12 INSPIRE Innovation in Sensor Programming Implementat
88. sensor network Wireless sensor networks are a subtype of ad hoc networks which are well stud ied However from a design standpoint wireless sensor networks are very unique compared to standard networks The nodes are potentially mobile and or unreliable and we need to approach these networks from a totally fresh viewpoint in order to analyze them Akyildiz et al 1 highlighted some of the main differences between wireless and traditional ad hoc networks in a list similar to the following The number of nodes in a WSN may be several orders of magnitude higher than in traditional networks Sensor nodes are prone to failure Sensor nodes are limited in energy computational capacities and memory Consequently WSNs cannot afford table driven MATNET protocols requiring too much memory to store routing tables As opposed to the one to many broadcast model common in ad hoc networks WSNs often use a many to one communication model with the topology of the reverse multicast tree Sensor nodes are densely deployed The topology of a WSN changes frequently Sensor nodes may lack global ID Because of these differences studies on WSN have developed into a seperate and distinct domain from conventional network development WSN Protocol Stack Akyildiz et al presented a protocol stack for WSN 1 which they modeled after the ISO OSI International Organization for Standardization open system interconnection model It has five
89. sensor networks the design of algo rithms becomes an important issue as energy and computational resources are scarce and therefore must be effectively put to good use With the limited computational ability of each individual node multiple sensors nodes collaborate to solve tasks us ing complex parallel processing techniques These parallel processing techniques rely on efficient parallel algorithms to achieve collaboration Therefore in this context a parallel algorithm can be defined as an algorithm in which several computations are carried on simultaneously across multiple processing units In order to extend the life of a sensor network these parallel algorithms have to be developed to be efficient in network resource usage hence the need for energy aware networking algorithms 7 10 9 There are numerous networked computing devices of all shapes and sizes from handheld computers such as personal digital assistants and mobile phones to more powerful systems such as laptops desktop workstations and super computers Most of these devices communicate over traditional IP based networks supporting huge data transfer rates due to rapidly increasing band width and faster processing capabilities These networks being highly scalable and structured may consist of several routers switches and bridges interconnecting millions of nodes and may use complex routing schemes to transfer data from one end device to another With energy and computationa
90. service provider Application layer support sublayer framework ZigBee device object Network layer Datalink layer Physical layer Most sensor applications based on ZigBee specifications typically interact with the application and network layers specified by ZigBee The network layer provides necessary abstractions for managing multihop communication between the various nodes while the application support layer manages communication between the var ious application objects Also the ZigBee alliance makes provisions for security Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 207 www finebook ir 208 STANDARDS FOR BUILDING WIRELESS SENSOR NETWORK APPLICATIONS WWAN WMAN IEEE 802 20 E WiMax IEEE 802 16 E WLAN Zigbee 802 15 3 wean 802 15 4 802 15 3a Bluetooth 802 15 1 802 15 3c 0 01 0 1 1 10 100 1000 Data Rate Mbps FIGURE 11 1 Comparison of IEEE 802 11 standards services device management and the ability to extend the application framework with vendor specific services The full architecture of the ZigBee stack is shown in Fig 11 2 1L2 ZigBee DEVICES AND COMPONENTS The maintenance of the 802 15 4 sensor network is the backbone of the easily pro grammable model of the network stack design As by design they are self organizing
91. specified layers are usually through abstract libraries implemented by independent vendors and offered through specific sensor operating systems e g TinyOS These libraries automatically transform application data into a form in conformance with IEEE specifications These vendor provided libraries providing software abstraction for the MAC and physical layers are referred to as the network layer In this section the several components that form the networking stack in wireless sensor networks are examined see also Fig 8 4 8 3 1 Physical Layer The physical layer is the first and lowest layer consisting of basic hardware transmis sion technologies of a network It is responsible for several functions including Provision of a data transmission service Management of RF transceiver Channel selection www finebook ir SENSOR NETWORK STACK 137 Energy and signal management functions For wireless sensor networks the phys ical layer transmits in one of three unlicensed frequency bands In North America the most common is the 915 MHz ISM instrument scientific medical band Among the main functions of the physical layer is the detection and correction of transmission errors These errors could stem from several factors such as Attenuation a decrease in intensity of electromagnetic energy at receiver due to long distance Doppler shift a change in frequency of a wave caused by the relative velocities of the transmitter an
92. such programs is a cumbersome activity In an application centric view users express data fusion and integration needs by describing relationships among objects in the domain of the control application Application centric views can be supported with any level of abstraction i e low medium or high However wireless sensor applications viewed with low level abstraction tend to be more conducive to a collaborative approach due to resource constraints than would a traditional software program In event driven programming you deal with concepts such as inversion of control blocking versus nonblocking code and run to completion RTC execution semantics Polling data are generated by the microframework which allows global tracking of the current hardware clock Once the poll is active the sensor object changes states if idle it will accept the poll and initialize itself with a predetermined timeout The timeout event is handled and the measured value from the sensor is read accurately and processed If there is a threshold set for this particular process then an alarm will be enabled according to the current read value This alarm event will be broadcasted to reach its destination using source destination pairs The on idle event can support the low power features of the target hardware enabling further energy savings www finebook ir DETAILS ON EMBEDDED DATA STRUCTURES 51 4 4 DETAILS ON EMBEDDED DATA STRUCTURES We now introduce the
93. that K is a regular graph of degree n 1 This is illustrated in Figs 4 16 and 4 17 4 7 3 Walk Path Cycle Let G V E be a simple graph Let vj v2 v3 v4 vy be some vertices of G and let vj be adjacent to v 1 i k We say that this sequence v v2 v is a walk from v to v if no edge appears more than once in the sequence v is called the starting point and u is called the terminus In the preceding definition of a walk some v and V may be the same for distinct i and j Consider a walk vi v2 Vk We say that the walk starts from v travels through vi v2 vg and finally reaches vg If vy vz the walk is called a closed walk A walk that is not closed is said to be open A walk in which no vertex appears more than once is called a path A closed www finebook ir GRAPH AND TREES 63 FIGURE 4 16 A regular graph of degree 2 walk in which no vertex appears more than once is called a cycle or circuit In other words a circuit is a closed path In Fig 4 18 a b c d b e f is a walk It is not a path because b is repeated a b c d is a path from a to d Also note that c b e p h e b d is neither a path nor a walk It is not a walk because the edge b e appears twice in the sequence once from b to e and then from e to b The sequence a b e f g a is a cycle In a cycle an edge joining two nonconsecutive vertices is called a chord In the cycle a b e f a in Fig 4 18 b f is a chord of
94. that a client needs to solve before asking to establish a con nection In this way a node is prevented from generating many extraneous requests at once 16 3 4 Transport Layer A desynchronization attack occurs when a malicious attacker uses falsified messages to cause two nodes to believe that they are out of sync and repeatedly run a syn chronization recovery protocol If the packets being received at the endpoints can be authenticated this attack will not work 16 3 5 Application Layer Attacks at this layer attempt to disrupt or crash individual services running on the target computer denying service in that way Many traditional hacking methods rely on exploiting software vulner abilities in this manner www finebook ir 302 SECURITY IN SENSOR NETWORKS 16 3 6 Conclusion Many of the aformentioned attacks could be prevented by using encryption to authen ticate data packets Unfortunately the limitations of small sensor nodes and those of ad hoc sensor networks in general make such approaches largely unfeasible As in most programming problems security threats to sensor networks can be handled most efficiently if they are considered during design so security aware design is very important 16 4 SOME WELL KNOWN ALGORITHMS FOR SECURITY PROBLEMS The following paragraph describes an overview of some of the known algorithms in the security aspects of sensor networks For the purpose of generality we are providing only an o
95. the cycle a b e f g a 4 7 4 Subgraph Let G V E be a graph Let V be a subset of V and F a subset of E such that all the edges in incident only on vertices of V G V E is called a subset of G The graph shown in Fig 4 19b is a subgraph of the graph in Fig 4 19a In a graph G for any two vertices u and v if there is a path from u to v the graph G is called a connected graph In a disconnected graph a maximal connected subgraph is called a connected component or simply a component Let V be a subset of V Let be the collection of all edges of G which have both end vertices in V Then X K 4 Ks FIGURE 4 17 Some completed graphs www finebook ir 64 DATA STRUCTURES FOR SENSOR COMPUTING FIGURE 4 18 A graph illustrating paths and walks G V G is called the induced subgraph of G induced by V The graph in Fig 4 19b is not an induced subgraph of Fig 4 19a The graphs in Figs 4 19c and 4 19d are induced subgraph induced by 1 5 4 and 2 3 4 respectively Two subgraphs are said to be edge disjoint if they have no common edge The graphs shown in Figs 4 b 2 BW b i 5 4 4 c d FIGURE 4 19 a A graph G b a subgraph G c subgraph induced by 1 4 5 subgraph induced by 2 3 4 www finebook ir GRAPH AND TREES 65 4 19c and 4 19d are edge disjoint subgraphs of Fig 4 19a The following observations can now be made 1 Any graph G is a subgraph of G itself 2 Any vertex i
96. the network 2 Every ZigBee network must contain a network coordinator Other full function devices FFDs may be found in the network and these devices support all of the 802 15 4 functions They can serve as network coordinators network routers or devices that interact with the physical world The final device found in these networks is the reduced function device RFD which usually serves only as a device that interacts with the physical world An example www finebook ir ZigBee APPLICATION DEVELOPMENT 211 Mesh Network gt Star Network Topology Intermediate Router FFD e Zigbee End Device s gt Zigbee Coordinator FIGURE 11 3 A typical ZigBee network of a ZigBee network is shown in Fig 11 3 The components needed to develop a real time automation application are 1 Coordinator node with network configuration and application 2 Routers that are in the form of clusterheads and allow multihopping 3 End devices connected to sensors The ZigBee PAN coordinator which is currently the most powerful node on the net work is normally 1 hop Furthermore the Wireless Sensor Network WSN commu nicates to a fixed Local Area Network LAN which has a gateway Internet Protocol IP address to transfer data serially or by built in ethernet ports in some models of ZigBee Sometimes a Sniffer program from an IP can be used to monitor the unknown topology of a WSN network Programming in the air can be used to correct some
97. time than a join and does not erase the neighbor table A join erases the neighbor table forcing all of the router s childen to re issue joins EVB LED1 OFF not connected to a network www finebook ir MOTIVATION AND BACKGROUND 229 aplRejoinNetwork routerState ROUTER STATE REJOIN WAIT break case ROUTER STATE REJOIN WAIT if apsBusy break if aplGetStatus LRWPAN STATUS SUCCESS my_timer halGetMACTimer routerState ROUTER_STATE_NORMAL ping_cnt 0 failures 0 EVB_LED1_ON conPrintROMString Network WHEREVER RON C Rejoin succeeded in printJoin Info routerState ROUTER STATE NORMAL j else failures if failures MAX REJOIN FAILURES conPrintROMString Network Rejoin Mi endl I max tries exceeded Trying a join routerState ROUTER STATE JOIN NETWORK j else conPrintROMString Network ReJoin FAILED E EEE A ei ro Ae cut EAEE PEN AA ATE TET Waiting _then trying again n my_timer halGetMACTimer wait for 2 seconds while halMACTimerNowDelta my timer lt MSECS_TO_MACTICKS 2 1000 routerState ROUTER STATE REJOIN NETWORK break j j break default routerState ROUTER STATE JOIN NETWORK end while 1 else conPrintROMString This application EEEE ei EAEE EATE CS is intended for a router WTn conPrintROMString Entering infinite UT EH loop doing nothing n while 1 endif www fineb
98. to node C Also worthy of note is the fact that RTS and CTS will have the MAC addresses of the sender and the receiver Further RTS CTS and the data packets will all have the dura tion T that it will take for the complete transaction Similarly a frame will indicate whether the transmission is unicast or broadcast The handshake protocol proceeds as follows 1 Node B senses the medium and goes to the next step when the medium is free 2 Node B broadcasts an RTS Note that it is a broadcast and not a unicast Also RTS contains the MAC addresses of B and C 3 Nodes C and A hear it The RTS contains the total duration d up to B receiving an acknowledgment www finebook ir 150 ALGORITHMS FOR WIRELESS SENSOR NETWORKS FIGURE 8 8 RTS CTS handshake configuration 4 Node C realizes that the RTS is for itself senses the channel acquires it and broadcasts a CTS This also contains the duration d Concurrently when A hears the RTS it realizes that it is not for itself but it will set the NAV flag ON to indicate that the medium is busy for the duration d1 5 When the CTS from C reaches B B unicasts one packet of data to C Note This unicast frame will contain the sender s and receiver s addresses and also a flat stating that the frame is a unicast frame Concurrently when D hears the CTS broadcast from C it will set its NAV flag ON for the duration d 6 Node C receives the data from B and then sends a unicast
99. transmit a packet This counter is incremented every time a periodic clock fires Assume that clocks are synchronized for this implementation 8 6 CARRIER SENSING VERSUS DECODING In Fig 8 7 the green circle shows the decode range and the yellow circle shows the carrier sense Note that these circles are part of the manufacturer s specs To decode the signal must be strong that is the signal to noise ratio SNR must be high To carrier sense the signal can be weak that is SNR can below We will now write programs for nodes B C D E F and measure the relative signal strengths Verify with the manufacturer s specs that is yellow circle and green circle www finebook ir CARRIER SENSING VERSUS DECODING 149 FIGURE 8 7 Illustration of carrier sensing versus of decoding begin if senseCarrier equals NIL print Medium is carrier free elif senseCarrier equals WEAK print Medium is carrier sense elif senseCarrier equals STRONG print Medium_is_carrier decode end If 8 6 1 RTS CTS Handshake Ready to send RTS and clear to send CTS messages are broadcast control bytes that are exchanged between the transmitter and the receiver to coordinate the trans mission of data between them A typical handshake protocol that uses this is shown in Fig 8 8 In this example we also show how to convert a timing diagram into programs Node B wants to transmit a data packet
100. use it is supported by numerous operating systems and sensing modules including TinyOS Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright 2011 John Wiley amp Sons Inc 27 www finebook ir 28 SENSOR TECHNOLOGY MOTE LAYER XMesh Sensor Apps SERVER LAYER Database Logger CLINET LAYER Visualization Analysis Tool PC Terminal MoteView as localhost Sensor Mesh Network PC MoteView Client Mirror Local Server Stargate Remote Server Stargate Emj PC MoteView Client FIGURE 3 1 An illustation of the various distinct categories of sensor tools Mantis OS and Contiki It includes the MicaZ Fig 3 2 Mica2 Cricket Fig 3 3 and Mica2Dot Fig 3 4 series of sensors Each of these motes has unique physical and functional capabilities which serve as a key differentiator Table 3 1 summarizes their similarities and differences effectively To extend the functionality of these motes additional sensor data acquisition boards can be attached using the expansion connector provided on the Mica motes FIGURE 3 2 A MicaZ sensor www finebook ir SENSOR LEVEL 29 FIGURE 3 3 A Mica2 sensor board FIGURE 3 4 A Mica2Dot sensor TABLE 3 1 Mica Family of Sensors Property Mica
101. visualization 12 num2str bags lost 1 x 1 getdata line 4 file event type end end disp Flight getdata line 4 file departed write to global db Bag flight BagAttribute www finebook ir SOURCE CODE 291 case 08 flight arrived disp Flight getdata line 4 file arrived tsensor visualization 11 0 0 0 getdata line 4 file event_type sdownload from global db Bag flight BagAttribute case 09 Photograph taken disp RFID for Bag getdata line 2 file _returned photo captured sensor visualization 14 getdata line 2 file 0 NOVO event type 2 capture photo here 2 remove refrences to bag getdata line 2 file from local db 2 save in natl db case 10 Transfer red to Lost Baggage area results 0 for x 1 line if strcmp getdata x 2 file getdata line 2 file amp amp strcmp getdata x 3 file 10 results results 1 end end sensor visualization 15 getdata line 2 file 0 NOVO event type if mod results 2 disp Bag getdata line 2 file went into lost baggage area into lost baggage area else disp Bag 4 getdata line 2 file went out of lost baggage area out of lost baggage area end otherwise disp ERROR INVALID EVENT ON LINE num2str line end sensor visualization m function sensor visualization LOCATION BAGI
102. void command uint32 t get t0 command void command uint32_t getdt This is particularly useful in the context of wireless sensor programming where we have to be aware of time and this is achieved by starting a clock at the beginning and managing it from then on Since batteries have limited lifetime we need to save energy by avoiding unnecessary transmissions particularly when other nodes are not listening Thus there is a need to go into sleep mode as often as possible and when the sensor node wants to go into sleep mode it has to make sure that other nodes are not trying to interact with it by for example trying to send some data or waiting for data to be sent This requires planning activities well in advance so that everything is sched uled and executed in the planned way and we need a timer for this In fact there are two important resources that a node must share with the other nodes time and the medium The Timer interface provides all basic operations that we need to manage the time within a node The Timer is also an example of parameterized interfaces In the Timer interface declaration precision tag is a type parameter that can be instantiated to give the timer the required precision For example TinyOS defines the following types typedef struct int notUsed TMilli typedef struct int notUsed T32khz typedef struct int notUsed TMicro www finebook ir 118 SENSOR PROGRAMMING These types define the pr
103. when the previous call continues it will find the value of x to be 3 erroneous and the function will return FALSE instead of TRUE One traditional solution to this problem is to make sure that when variables such as those stated above are shared across multiple processes they must be protected and this is done in nesC by declaring atomic blocks of code 6 2 4 Atomic Block An atomic block of statements is a sequence of statements that are executed to completion without preemption being interrupted by the TinyOS system When async functions are used we need to be careful about variables shared among multiple components In such situations atomic blocks of statements are useful On the other hand excessive use of atomic blocks of statements can degrade the performance of the TinyOS system Finally care must be taken to ensure that cyclic dependences do not occur in atomic code since they lead to deadlocks in the system Thus programs using async commands must make use of atomic blocks only when shared variables are used An example of an atomic statement is presented below int x 1 async command int add atomic x x1 if x 2 0 return TRUE else return FALSE In this example multiple invocation of the async command share the values of the variable x However for a given invocation the atomic block is not preemptable and thus we will clearly know what result this function is returning Note that a
104. www finebook ir A SIMPLE PROGRAM 101 else err FAIL signal PrintMessage sendDone msg err PrintClientC This is a client module that uses a set of interfaces to print a Hello World message In this case it uses two interfaces Boot and PrintMes sage and thus provides implementation for all the events defined in these two interfaces The nesC code for the Boot interface is shown below interface Boot event void booted The Boot interface is synonymous with the main function in traditional C programs providing a point from which program execution starts that is by first initializing all components necessary for successful program execution module PrintClientC uses interface Boot uses interface PrintMessage implementation event void Boot booted call PrintMessage send Hello World 11 event void PrintMessage sendDone char msg 20 error t err if err SUCCESS dbg std SendDone signalled with nowerrors n else dbg std SendDone signalled with an error n So far we have defined an interface and implemented all our commands in one module and the events in another module Although our program consisting of two interfaces and two modules is logically ready to run there is one www finebook ir 102 PROGRAMMING IN NesC more issue that must be resolved before the program can execute successfully The PrintClientC module uses two interfaces Boot PrintMessage
105. www finebook ir PREFACB Xv development may skip to Chapter 7 Sensor Programming In conjunction with the material in the book readers are encouraged to work on moderate sized projects taken from the book S S IYENGAR Louisiana State University Baton Rouge USA N PARAMESHWARAN University of New South Wales Sydney Australia V V PHOHA Louisiana Tech University Ruston USA N BALAKRISHNAN Indian Institute of Science Bangalore India C D OKOYE Louisiana Tech University Ruston USA January 28 2010 www finebook ir Foreword Dr Iyengar et al have written an excellent piece of senior undergraduate or first year graduate level text for sensor or sensor network programming The book contains numerous practical examples of real or pseudocode and will be extremely beneficial for both students and teachers It will also be useful for working engineers writing code for sensor based complex or simple systems Lately the use of sensors to measure space or temperature has grown exponentially from freeways to airports to medical devices to smartphones Dr Iyengar et al s book will have a positive impact to the industry in general Arup Gupta Director Wireless Platform Technologies Ultra Mobile Group Intel Corp xvii www finebook ir Acknowledgments This book started out with a series of discussions in 2008 and evolved from various research projects on sensor networks It was contributed to and
106. 07 Index 313 www finebook ir Preface The price of greatness is responsibility Winston Churchill Sensor processing is a central and an important problem in aerospace defense au tomation medical imaging and robotics to name only a few areas A surveillance system used in aerospace and defense is an example of a sensor processing system It uses devices such as infrared sensors microwave radars and laser radars that are capable of detecting and tracking flying objects in their observational space A sensor processing system may employ intelligent and disparate sensors that are distributed logically spatially and even geographically It is then referred to as a distributed sensor network DSN The sensor may measure scalar values e g temperature or vector values e g position in three dimensional space The measurements are gen erally a function of time and or space Because of variation in operating environments or other factors such as aging and communication delays the measurements may appear contradictory Although combining the numerous sensor measurements may appear contradictory it minimizes the uncertainty of measurements and improves reliability and fault tolerance There is a wide body of literature on sensor networks and on design analysis protocols and other research related issues in sensor networks However the issues of software development in particular pedagogical material related to software de ve
107. 150 161 Kumar R and M Fabian Supervisory control of partial specification arising in protocol conversion Proc 35th Allerton Conf Communication Control and Computing 1997 pp 543 552 www finebook ir 310 BIBLIOGRAPHY Kumar R and V Garg Modeling and Control Logical Discrete Event Systems Kluwer Academic 1995 Kumar S and E H Spafford A generic virus scanner in C Proc 8th Computer Security Applications Conf 1992 Le V T D Creighton and S Nahavandi Simulation based input loading condition optimi sation of airport baggage handling systems Proc IEEE Intelligent Transportation Systems Conf Seattle WA 2007 LeCharlier B and M Swimmer Dynamic detection and classification of computer viruses using general behavior patterns Proceedings of 5th Int Virus Bulletin Conf Sept 1995 p 75 Lee W and S J Stolfo Data mining approaches for intrusion detection Proc 7th USENIX Security Symp SECURITY 96 Jan 1998 Leone K and R Liu The key design parameters of checked baggage security screening systems in airports J Air Transport Mgmt 11 69 78 2005 Levin R B The Computer Virus Handbook Osborne McGraw Hill 1990 Levis P and D Gay TinyOs Programming Cambridge Univ Press 2009 Linz P An Introduction to Formal Languages and Automata 3rd ed Jones amp Barlett Oct 2000 LSU Research Group LSU Sensor Simulator LSU SenSim version 1 Jan 2005 User Manual Dept Comput
108. 3 Most references to the term sensor network can denote multiple sensing configura tions to be used in multiple contexts Sensor networks typically consist of numerous sensing devices that may communicate over wired or wireless media and may have as intrinsic properties limitations in computational capability communication or energy reserve This does not imply that all sensor deployments consist of severely resource constrained devices for example radar closed circuit cameras and other wireline devices are commonly used in sensor network experimentations in academia and military research These sensing devices possess reasonable computational capabil ity and more importantly may not have limited energy or constrained communication www finebook ir NEXT GENERATION SENSOR NETWORKED TINY DEVICES 5 abilities The main crux of this book is focused on the class of sensors having severely constrained computation communication and energy resources These devices range from penny to matchbox in size and are deployed in an ad hoc and nonplanned ran dom fashion Examples of such devices include the mote platforms commonly used in academia 12 NEXT GENERATION SENSOR NETWORKED TINY DEVICES 1 21 Domain Specific Challenges Development of software in wireless sensor networks draws on experiences across several domains in computer science and some engineering disciplines such as 1 Networking Networking knowledge is critical in sensor netw
109. 8 offset FAILED fontsize 10 else text DISPLOC 2 50 DISPLOC 1 6 18 offset PASSED fontsize 10 end if viz bag loc data i 5 6 if strcmp viz bag loc data i 4 NOT SCANNED text DISPLOC 2 50 DISPLOC 1 6 18 offset WARNING Bag 8 viz bag_loc_data i 2 SES RACE Scania L SCR acad paci ci A i loaded onto flight num2str viz bag loc data i 3 never passed security inspection fontsize 10 for tmp 1 3 pause 2 beep end else www finebook ir text DISPLOC 2 50 DISPLOC 1 6 18 offset fontsize 10 end if viz bag loc data i 5 7 text DISPLOC 2 10 DISPLOC 1 20 flight_ viz_bag_loc_dataf i 4 fontsize 10 end 4esefzssweeseceucueecueiee pum eem if viz bag loc_data i 5 8 not visualizing right now end end sensor visualization clear m function sensor visualization clear global DISPIMG global viz bag loc data global Environment if size DISPIMG 1 Environment imread maindiagram jpg jpg Environment rgb2gray Environment end DISPIMG Environment viz bag loc data l PROBLEMS PROBLEMS PASSED 295 15 1 Design an event generator that generates synthetic testing data for the BHS 15 2 Propose a sensor network configuration applicable to geoscience prob lems that can also be extended to the social sciences or people centric situations 15 3 Describe an example integrat
110. A node such as B will attempt to answer this query but in this case since B has no knowledge of the event it passes the query to its neighbors Eventually the query reaches node j which generates an answer to www finebook ir 180 TECHNIQUES FOR PROTOCOL PROGRAMMING Source Node Q Destination Node Q Intermediate Node FIGURE 9 2 Flooding dataflow the query and sends it to its neighbor node k for example with a request to forward it to node c Now node k will store i e cache a copy of the reply and forward it to one of its neighbors This process continues until the reply reaches node c If this query is initiated at any time in the future the intermediate nodes that previously cached a copy of node j s reply in their local memory can generate a reply to the query initiator CONTENT BASED ADDRESSING self initial Value Sink role if self sink then flood network with request establishing path to sink if self sink amp reply received then accept it source role if self source then send reply along path intermediate role if self intermediate node on the path amp www finebook ir FLOODING 181 have not already forwarded the request then forward the request and record the path i if self intermediate node on the path amp have not already forwarded the answer then forward the answer and store the answer booted self initialValue if self
111. ABLE 15 1 Preexisting Data File in BHS Simulation Event Type Description Col2 Col4 Col5 01 Checkin Bag ID Flight ID 02 Weighin Bag ID Weight Ib 03 Security scan Bag ID 0 1 success fail 04 In out of storage Bag ID Airport storage area ID 05 In transfer van Bag ID Van ID 06 In plane Bag ID Flight ID 07 Flight departs Flight ID 08 Flight arrives Flight ID 09 Checkout Bag ID Airport ID Carrier 10 In out of lost baggage Bag ID Airport ID Carrier Our simulation computes the following performance metrics of the BHS Percentage bags that missed fiight Average bag time to storage Average bag time on loop i e time in queue The performance metrics obtained for the test data file are shown in Table 15 3 Vis Event Handler Visualization FIGURE 15 7 Simulation engine www finebook ir asesseq 1so Jo 1no ul OI asequiep eqoys 03 K1onb Jo sj nso1 ased pue yng Z eaepreoor zpre17 moyo 60 oloud exer eni moyo 60 osequyep LInsaa ur en eA Aue Z pTety B90 o amp ionb jo sj nso1 ojsed pue yng 90 DT TjJ BB v eaepreoor vpI917 SAIE UFA 80 asequiep 0 gp eoo ased pue m asequyep Z 1unssw Aue Z P1913 eqo 8 03 K1onb Jo sj nso1 ased pue yng 90 DT TJ 2 v e3epreo2or vpT9T13 syredop 1431A LO JSOL 0 Seq JO 9371s JOS o1oqAouios WY SOT 0 lt 90 E
112. AC layer If the network packet length is more than that of the MAC frame it is fragmented and the fragments for that network packet are created with MAC headers and are inserted into the fragments queue The MAC layer then waits for the channel to be idle to send its frame from the messages queue The MAC layer has a NAV timer which specifies the busy idle state of the medium The NAV timer set for a node implies that the channel is busy When the NAV timer expires the MAC layer waits for the channel to be free for DIFS time and if the channel is still idle after DIFS timer has expired it then goes into exponential backoff It then waits for a random time set by the backoff timer The backoff timer decrements its value during the idle period of channel The node whose backoff timer expires earlier will get the chance to transmit its next frame All the intermediate nodes receive this frame and set their NAV timers to the values obtained from the header field of the received frame Then the backoff timer of the intermediate nodes is stopped from decrementing Once the channel becomes idle when the NAV timer expires all the nodes start decrementing their backoff timers The node whose backoff timer expired earlier and got the channel will send the first message from the messages queue If it is a broadcast message then all the nodes in its region receive it and the MAC layers of those nodes encapsulate the network packet and send it to the network la
113. AM J Control and Optim 25 3 206 230 1987 Research Integration Platform Survey embedded WiSeNts consortium Rhee L A Warrier M Aia J Min and M L Sichitiu Z mac A Hybrid MAC for Wireless Sensor Networks 2008 IEEE Press Piscataway NJ vol 16 pp 511 524 www finebook ir BIBLIOGRAPHY 311 Rijsenbrij J C and J A Ottjes New developments in airport baggage handling systems Transport Plann Technol 30 417 430 2007 Ruiz Sandoval M T Nagayama and B F Spencer Sensor development using Berkeley mote platform J Earthquake Eng 10 289 309 2006 Sastry S Smart space for automation Assembly Autom 24 2 201 209 2004 Sastry S S S Iyengarand and N Balakrishnan Sensor technologies for future automation systems Sensor Lett 2 1 9 17 2004 Sastry S and S S Iyengar Distributed Sensor Networks CRC Press 2005 Sastry S and S S Iyengar A Taxonomy of Distributed Sensor Networks CRC Press 1995 Schultz M G and E Eskin et al Data mining methods for detection of new malicious executables Proc IEEE Symp Security and Privacy May 2001 Sobieh A and J C Hou A Simulation Framework for Sensor Networks in j sim Technical Re port UIUCDCS R2003 2386 Dept Computer Science Univ Illinois Urbana Champaign Nov 2003 Solomon A and T Kay Dr Solomon s PC Anti virus Book Newtech 1994 Spinellis D Trace A tool for logging operating system call transacti
114. CAST ADDR amp msg sizeof sensorData call Leds ledOToggle We will assume that there are no more than 30 motes in the neighborhood at any given time The mote address of the given node where we want to build the neighborhood table is 0 its MAC address is 0x001EDD32 and the identity of the group which consists of the current node and all its neighbors is 1 We begin the table construction by starting the radio www finebook ir 142 ALGORITHMS FOR WIRELESS SENSOR NETWORKS Starting the Radio Component We set the group address to 1 and the current node address to 0 and then start a monoshot clock that invokes the event function Timer0 fired where we broadcast our node address to all neighbors and wait for their replies 8 4 3 Implementation Implementation Details of the Receive Method As soon as we receive a message from any neighbor say P we check whether it is a reply to the message that we sent and if it is we store P s ID in our table However if the message is a request asking us to send our ID we then send our ID to node P To achieve these operations we use the following functions sendSensorInformation and populateSen sorTable event message t Receive receive message t msg void payload uint8 t len sensorInfo myInfo Specified in header file myInfo sensorInfo payload if myInfo gt requestCode 0x1 call Leds led2Toggle Toggle Led 3 to indicate info receive
115. D number special event type global DISPIMG global viz bag loc data global Environment global time Environment imread maindiagram jpg jpg Environment rgb2gray Environment Init coordinates Deski 21 240 1 Desk2 21 338 2 Desk3 21 440 3 Scannerl 326 157 4 Scanner2 299 533 5 www finebook ir 292 MATLAB SIMULATION OF AIRPORT BAGGAGE HANDLING SYSTEM Scanner3 440 631 6 XRAY1 218 257 7 XRAY2 218 322 8 LEVEL4INS 173 659 9 STORAGE 480 590 10 PLANE 540 100 11 LOSTAREA 577 728 12 VANLOC 550 300 13 PICKUPAREA 150 690 14 LOSTBAGGAREA 465 690 15 Initializing coordinates LOCATION 4 BAGID 006 if exist DISPIMG DISPIMG DISPIMG Environment end if exist viz bag loc data viz bag loc data end viz bag loc data size viz bag loc data 1 41 1 LOCATION 2 BAGID 3 number 4 special viz bag loc data size vizbag_loc_data 1 viz_bag_loc_data size viz_bag_loc_data 1 viz bag loc data size vizbag_loc_data 1 D D D viz bag loc data size viz bag loc data 1 5 event type switch LOCATION case DISPLOC Desk1 case 2 DISPLOC Desk2 case 3 DISPLOC Desk3 case 4 DISPLOC Scanner1 case 5 DISPLOC Scanner2 case 6 DISPLOC Scanner3 www finebook ir SOURCE CODE 293 case 7 DISPLOC XRAY1 DISPLOC XRAY2 DISPLOC LEVEL4 INS case 10 DISPLOC STORAGE DISPLOC PLA
116. D E can be obtained only by moving the rest of the data items in the list by appropriately changing the pointers Another problem is that of deleting lists It is awkward to keep track of free space within this large array Thus an effective implementation for the maintenance of several lists is not to be found using linear lists 4 5 1 Examples of Linked Lists An alternative data structure to linear lists that allows for a greater amount of flexibility is the linked list List elements are called nodes A node consists of data and a link to another node in the list The beginning of the list is indicated by a pointer The name of the list is actually a pointer to the first element in the list A linked list structure for our previous example is given in Table 4 6 The pointer is set to O if it would take any value that cannot be used as a link or if no link exists If we extract the first second and third lists we could draw it as shown in Fig 4 8 The name of the list structure is one In particular the lists can grow to arbitrary length We may be able to keep several lists simultaneously in one area of storage with the only space restriction that their total storage utilization will be less than or equal to the area available There is an obvious disadvantage to linked lists in terms of searching If a linear list is kept in sorted order then it may be searched with a binary search Consider TABLE 4 5 Linked List Data Structure Name P
117. E 13 7 Translucentlayering in service implementation O O O Q O e 40 O O a 2s Sampling Protocol Slotted Protocol FIGURE 13 8 Clustering data aggregation Application Protocol layer4 Transport Protocol UT Network Prorocol Header e g TCP Prorocol Handler e g UDP Per node Layer diceaiches gar Protocol layer2 MAC Protocol layer1 Physical Power i Sensor Physical Layer FIGURE 13 9 GlomoSIM architecture www finebook ir SERVICE ARCHITECTURE 245 Current STACK Modified STACK GlomoSIM Code Generator Scipts r 1 T Application GlomoSIM App Hardware Abstraction Hardware Interfaces Hardware GlomoSIM Radio FIGURE 13 10 Application code development cycle due to the node deployment density ratio The key elements of the service protocol design are listed in Tables 13 1 and 13 2 13 2 1 Reliability The traditional concept of link abstraction does not hold true for WSN as it not only communicates wirelessly but also performs data sensing which is prone to outside spurious noise and channel interference Due to this design constraint the link quality is defined to detect good data from occurring noise levels This helps application developers rely on the data being measured and validates the significance of the data to the deployed applic
118. FRL and the Indian Institute of Science Bangalore Dr Parameshwaran acknowledges the support and assistance needed from University of New South Wales Sydney Australia during his sabbatical at LSU The authors also acknowledge Prof Balakrishnan s research group at Indian Institure of Science Bangalore during the preparation of this manuscript Prof Phoha acknowledges the support of LaTech researchers and many of his graduate students Kiran Balagani Chuka Okoye Sunil Babu and colleagues Enam Karim and Rastko Selmic from Louisiana Tech University We would also like to thank Prof Holger Karl and Dr Adreas Willig the authors of the book Protocols and Architectures for Wireless Sensor Networks John Wiley amp Sons 2005 which was instrumental in providing the intial impetus for writing a book at a level with more programming details using nesC upon TinyOS Perhaps most importantly I would like to thank Dean Kevin Carman for his con tinued support and mentorship in my research activities at Louisiana State University www finebook ir About the Authors Each of the five authors has teaching and research interests in sensor networks A brief biographies of these five authors follow Dr S S Iyengar S S Iyengar is currently the Roy Paul Daniels Professor and Chairman of the Computer Science Department at Louisiana State University He heads the Wireless Sensor Networks Laboratory and the Robotics Research Laboratory at LSU
119. Fundamentals of Sensor Network Programming 6i ay re TA and it e A V S SITHARAMA IYENGAR NANDAN PARAMESHWARAN VIR V PHOHA N BALAKRISHNAN CHUKA D OKOYE www finebook ir FUNDAMENTALS OF SENSOR NETWORK PROGRAMMING www finebook ir FUNDAMENTALS OF SENSOR NETWORK PROGRAMMING Applications and Technology S Sitharama Iyengar Nandan Parameshwaran Vir V Phoha N Balakrishnan Chuka D Okoye IEEE WILEY A John Wiley amp Sons Inc Publication www finebook ir Copyright 2011 by John Wiley amp Sons Inc All rights reserved Published by John Wiley amp Sons Inc Hoboken New Jersey Published simultaneously in Canada No part of this publication may be reproduced stored in a retrieval system or transmitted in any form or by any means electronic mechanical photocopying recording scanning or otherwise except as permitted under Sections 107 or 108 of the 1976 United States Copyright Act without either the prior written permission of the Publisher or authorization through payment of the appropriate per copy fee to the Copyright Clearance Center Inc 222 Rosewood Drive Danvers MA 01923 978 750 8400 fax 978 750 4470 or on the web at www copyright com Requests to the Publisher for permission should be addressed to the Permissions Department John Wiley amp Sons Inc 111 River Street Hoboken NJ 07030 201 748 6011 fax 201 748 6008 or online at http www wiley c
120. GATION 201 if serial busy FALSE serial_packet call SerialPool get if serial packet NULL Out of memory so drop all packets error return msg else Prepare packet for sending operation uintl6 t sense_value uint16_t call SerialSend get Payload serial packet memcpy sense_ value uint16 t payload len if call SerialQueue enqueue serial packet FAIL error serialSuccess serial_busy TRUE post sendSerialData else If serial bus is busy we enqueue for now serial_packet call SerialPool get if serial packet NULL Out of memory drop all packets error return msg else uintl6 t queue value uintl6 t call SerialSend getPayload serial packet memcpy queue value uintl16 t payload len if call SerialQueue enqueue serial packet FAIL error return msg www finebook ir 202 SENSOR DEPLOYMENT ABSTRACTION event void SerialSend sendDone message t msg error_t err serial_busy FALSE if call SerialQueue empty FALSE Serial busy TRUE post sendSerialData task void sendSerialData buffer call SerialQueue dequeue if call SerialSend send Oxffff amp buffer sizeof message t FAIL error serial_busy TRUE void error call Leds ledOToggle void serialSuccess call Leds led1Toggle void radioSuccess
121. I reference 4 Running the Simulation We can run the simulation after all the layers are defined by configuring the parameters in the omnetpp ini file After writing your code at various layers do make in that corresponding subdirectory After make is done successfully return to the samples directory and change the parameters in the omentpp ini file The topology simulation time and other properties can be varied in this file These sections are clearly explained in the OMNeT manual Change the class name of the module you created as discussed in Section 14 3 and run the simulation You can redirect the out put to another text file Base Layer Concept The functionality of any layer of a node is defined in the Layer Base file This itself gets its properties from the SimpleModule of OMNeT All the base classes of the node are derived from LayerBase This is defined in the src directory Base Modules We provide base modules for each layer which inherits its properties from Layer Base It is implemented primarily to clearly define the interface that can be understood easily and that can be extended if necessary It implies that these source code files handle the basic functionalities of that layer and the user need not go through all this for the implementation Users only need to derive their class from these base modules These modules have header files in the inc directory The ned files and the source code files are ava
122. Installing TinyOS in Windows TinyOS is supported on the Windows platform through the use of Unix emulation software Cygwin The following steps describe how TinyOS libraries can be installed on a Windows workstation 1 Install the Java JDK The latest version can be downloaded from the Sun Java Website 2 Download and install Cygwin from http www cygwin com 3 Install native compilers for sensor motes The MSP430 toolchain for the Telos Tmote family of motes or the AVR tool chain for the Mica family can be downloaded from http www tinyos net dist 2 0 0 tools windows 4 Install the nesC compiler from the TinyOS Website 5 Finally install the TinyOS source tree that will enable you to compile and install TinyOS programs It can also be installed from the repository listed in step 4 An alternative Windows installation method is to install the MoteWorks package provided free from Crossbow Some more resources on Tinyos net available at http www tinyos net dist 2 1 0 tools windows describe how TinyOS can be in stalled and configured in Redhat Windows and other Debian based Linux distribu tions PROBLEMS 3 1 What are the three distinct categories of tools used for managing wireless sensor deployments 3 2 Using the procedures outlined in the programming tools install and configure your TinyOS programming environment www finebook ir 38 3 3 3 4 3 5 3 6 SENSOR TECHNOLOGY What does the term middleware re
123. It is for this reason that new low power components are being created and power saving policies such as dynamic power management DPM and dynamic voltage scaling DVS are continually being refined to provide ever increasing energy savings A typical sensor node is shown in Fig 8 2 www finebook ir 134 ALGORITHMS FOR WIRELESS SENSOR NETWORKS MMCX connector Atmel ATMegat 28 female External power connector X 51 pin Hirose connector On Off Switch malg FIGURE 8 2 A typical sensor node mote 8 2 DISTINCTIVE PROPERTIES OF WIRELESS SENSOR NETWORKS With the unpredictable nature and unsuitability of traditional routing algorithms in wireless sensor networks newer more efficient and more reliable algorithms have been developed These algorithms were developed to address the challenges posed by the volatile wireless communication in sensor networks These challenges stem from the ubiquitous and heterogeneous nature of WSNs making central management of individual sensor nodes nearly impossible As a result intelligent features such as self configuration healing dynamic routing and multihop communication abilities have been suggested to improve the reliability and management tasks in these networks In the following section we discuss each of these attributes and how they enhance communication in wireless sensor networks 8 2 1 Self Configuration Wireless sensor networks are usually composed of thousands
124. NE case 12 DISPLOC LOSTAREA case 13 DISPLOC VANLOC case 14 DISPLOC PICKUPAREA case 15 DISPLOC LOSTBAGGAREA otherwise null end for i DISPLOC 1 5 1 DISPLOC 1 5 for j DISPLOC 2 5 1 DISPLOC 2 5 DISPIMG i j 0 end end warning off imshow DISPIMG Struesize warning on for i 1 1 size viz bag loc data 1 LOCATION viz bag loc data i 1 switch LOCATION case 1 DISPLOC Desk1 Uu LOC Desk2 DISPLOC Desk3 DISPLOC Scanner1 DISPLOC Scanner2 DISPLOC Scanner3 DISPLOC XRAY1 DISPLOC XRAY2 www finebook ir 294 MATLAB SIMULATION OF AIRPORT BAGGAGE HANDLING SYSTEM case 9 DISPLOC LEVELAINS case 10 DISPLOC STORAGE case 11 DISPLOC PLANE case 12 DISPLOC LOSTAREA case 13 DISPLOC VANLOC case 14 DISPLOC PICKUPAREA case 15 DISPLOC LOSTBAGGAREA otherwise null end offset length find cell2mat viz bag loc data 1 i 1 LOCATION 1 text DISPLOC 2 10 DISPLOC 1 6 18 offset viz_bag_loc_data i 2 fontsize 10 text 600 10 datestr time fontsize 10 if viz bag loc data i 5 2 text DISPLOC 2 50 DISPLOC 1 6 18 offset num2str viz bag loc data i 3 lbs fontsize 10 if strcmp viz bag loc data i 4 overweight text DISPLOC 2 100 DISPLOC 1 6 4 18 offset OVERWEIGHT fontsize 10 if viz bag loc data i 5 3 if strcmp viz_bag_loc_data i 4 1 text DISPLOC 2 50 DISPLOC 1 6 1
125. ON For the sake of completeness in the context of programming this chapter provides a cursory view of security attacks and concerns in sensor network For more details on algorithms and architecture readers are advised to refer to some of the papers listed at the end of this chapter as well as the rich literature listed in the Bibliography and available elsewhere 16 2 SECURITY CONSTRAINTS The fundamental constraints under which sensor networks operate prohibits them from using public key encryption systems and third party authentication systems These constraints are described in the following subsections 16 2 1 Resource Constraints Resource constraints drive every aspect of sensor programming As noted the low power and processing capabilities of sensors are the most significant factors in sensor security A typical sensor node might have a maximum of around 20 30 J joules of energy For example a Berkeley mote has an 8 bit 4 MHz processor which supports a minimal reduced instruction set computer RISC like instruction set without support for multiplication or other costly operations Perrig et al 1 2 showed that a simple random structures algorithms RSA operation takes on the order oftens of seconds on this processor After a mote is loaded with the requisite os and communications software it has less than 4 kB of free space With space at such a high premium it is not possible to store too many long keys or even long algo
126. PacketSenderC continued event void Boot booted call PSControl start setLedBits 0x1 Packet was sent successfully event void PSControl startDone error_t error 1 if error SUCCESS call Timer0 startPeriodic TIMER FREQ else call PSControl start event void TimerO fired dbg x Timer Fired n setLedBits 0x5 0x5 0011 sendPacket void setLedBits uint8 t state call Leds set state void sendPacket BlinkData myBD counter j create a packet then send it myBD BlinkData call Packet getPayload amp pkt NULL myBD gt count counter if err call AMSend send AM_BROADCAST_ADDR amp pkt sizeof BlinkData dbg x AMSend reports success n busy TRUE else if err FAIL dbg x Error Could not send packet n www finebook ir PROBLEMS 129 event void PSControl stopDone error_t error does nothing for now We have not had any opportunity to invoke PSControl stop so in this imple mentation there was no need to invoke the event function PSControl stopDone A counter initialised to 0 has been used to provide the data in the packet being sent The dbg function prints messages on the command window if the program is run on the TOSSIM similator When we run this program on the sensor nodes the dbg output is ignored The boolean variable busy is to indicate whether the sender is busy or not mo
127. S Il xL S sebieyoos Jaye pejejduioo spunoi Hunnos ejeq 9 pajejduioo spunou unnos ejeq HO q peiejduioo spunoJ Bunnoi ezed HO 000 86 86 8289 69 8S Sh OE 000 86 86 8289 65 8S Sb OE 000 86 86 8289 6G 8S Gv OE r T Sy iei F T a 0 D 0 r T T T T T T T I n a 3 J 3 J Vu a Ja c Jo 1 702 E 109 3 J 5 joe 9 108 5 ne J 7 j o 70i o o jos 4 Orb L J m 1o gt J 5 09 3 Joa Q a ELS 002 5 J p a deels ym xy B ddy SW y4 sob eip a d ym xy 1 a ap XH x 2 oz Ov 09 08 oo ozo or 09 2 081 9 ooz amp o dwe saq 247 www finebook ir 248 PERFORMANCE ANALYSIS OF POWER AWARE ALGORITHMS study the effects of ultra low duty cycling applications The QoS application of the service can use the predefined data delivery 13 2 3 Minimal Application How to implement software in sensor networks is essential to understanding sensor networks A network architecture and protocols are essential foundations for building software applications 13 2 4 Data Routing This category of routing does not have any application control at the higher level As soon as the data are sampled they are forwarded to the nearest forward node to be delivered to the destination The performance of such an application is based on more efficiently forwarding the data toward their destination 13 2 5 A
128. SS error if call AMA amAddress 1 call RootControl setRoot you are root if call SerialControl start error else 199 in our code a serial packet Called whenever we send a radio packet Actual task to send data to computer If your ID is 1 SUCCESS call Timer0 startPeriodic periodic www finebook ir 200 SENSOR DEPLOYMENT ABSTRACTION event void SerialControl startDone error t err event void RadioControl startDone error_t err event void SerialControl stopDone error t err event void RadioControl stopDone error_t err event void LeafSend sendDone message_t msg error_t err async event void AMA changed event void Read readDone error_t err uintl6 t val if err FAIL sensor_value val error Timer will only be fired on leaf nodes When it happens we send a reading event void TimerO fired uint16_t data if call Read read FAIL error data uintl16 t call LeafSend getPayload amp buffer data sensor value if call LeafSend send amp buffer sizeof uint16_t SUCCESS radioSuccess This receive method is only signalled on the root node When it receives a packet it tries to send it to the serial interface event message_t Receive receive message t msg void payload uint8 t len message t serial packet First we check serial bus www finebook ir DATA AGGRE
129. T M25P80 flash serial ID CC2420 IEEE 802 15 4 radio FIGURE 3 7 A TelosB sensor 8 The functional and physical characteristics of both platforms are compared in Table 3 2 As in the Mica family of sensors additional data acquisition boards can be attached to these motes to allow for a more diverse sensing ability One such example is the bumblebee radar board Fig 3 9 which provides a pulsed Doppler radar for TelosB and Tmote sky sensors 3 1 3 Imote2 The Imote2 is a sensor platform built around the Intel PXA271 Xscale processor with a built in 2 4 GHz antenna 1 It is a powerful platform supporting computationally intensive tasks such as digital image processing due mostly to the scaling capabilities FIGURE 3 8 A Tmote sky sensor www finebook ir 32 SENSOR TECHNOLOGY TABLE 3 2 Telos Tmote Properties Specifications Telos Platform Tmote Sky Program flash memory kB 48 48 RAM kB 10 10 ROM kB 16 16 A D converter bits 12 12 Frequency band MHz 2400 2483 5 2400 2483 5 Data transmit rate kbps 250 250 Outdoor range m 75 100 50 125 Size mm 65x31x6 65x31x6 Light sensing range nm 320 720 320 720 Temperature range C 40 123 8 40 123 8 Humidity range RH 0 100 0 100 Percent relative humidity of its processor ranging from 13 to 416 MHz It has 32 MB of memory and supports data rates of lt 250 kbps It is currently supported by TinyOS and some Linux variants and can be order
130. TECHNIQUES FOR PROTOCOL PROGRAMMING 9 7 SELECTIVE REPEAT SELECTIVE REJECT The following program is somewhat simpler than the preceding one In this both the transmitter and the receiver have buffers The transmitter keeps sending its data packets while the receiver continues to receive them even if they are out of sequence The receivers ends J ACK for received packets and ACK for missing packets The transmitter then resends those packets for which v ACKs were received While this protocol is more efficient than the ones above the buffers that it uses will consume more memory in the sensor nodes Transmitter A buferl 0 7 0 pO 1 pl 2 p2 7 PT send now 0 7 0 1 7 send next 0 7 non empty junk j 1 end 0 initially start sending from bufferl while send next empty PARALLEL PARI for i 0 i lt size i il send now il send to node B bufferl i1 end 1 PAR2 while end 1 wait for ack from node B packet id pid extract control byte ack ack type extract ack type ack if ack type negative send next j pid end while send now send next size size of send next end PAR2 end PARALLEL end while www finebook ir SELECTIVE REPEAT SELECTIVE REJECT 169 receiver B buffer 0 7 empty i 05 while i lt 7 receive packet p fro
131. TWORK conPrintROMString Trying to is chee eee rejoin network Win aplRejoinNetwork rfdState RFD STATE REJOIN WAIT break RFD STATE REJOIN WAIT if stack is busy continue if apsBusy break if aplGetStatus LRWPAN STATUS SUCCESS failures 0 EVB LED1 ON conPrintROMString Network pene eret ches Rejoin succeeded n printJoinInfo rfdState RFD STATE NORMAL else failures if failures MAX REJOIN FAILURES this starts everything over www finebook ir 233 234 INSPIRE conPrintROMString Max Rejoins E TERM REN ae AOE failed trying to join n rfdState RFD_STATE JOIN_NETWORK else else wait to try again conPrintROMString Network Pe Be bee ett es Rejoin FAILED Re ia ater ages ea St a tiles qu Sx od Mogunt Waiting then trying again n most apps probably do not need to wait before retrying rejoin this is included just for visibility purposes in reading the output my_timer halGetMACTimer wait for 2 seconds while halMACTimerNowDelta my timer MSECS TO MACTICKS 2 1000 rfdState RFD STATE REJOIN NETWORK break endif default rfdState RFD STATE JOIN NETWORK else aplFormNetwork wait for finish while apsBusy apsFSM conPrintROMString Network formed CHAT RUN RE RR D E PORE usa EE kuupa aa au RPT waiting for RX n EVB_LED1_NO coordinator or rout
132. TinyOS 5 2 Distinguish between modules and configuration 5 3 For the car example illustrated in Figs 5 2 and 5 3 give nesC code to specify the car interface car configuration tire module engine module and car lighting module 5 4 Extend the nesC code in problem 5 3 by adding brake and engine modules to the car configuration and by adding the driver configuration to the car interface 5 5 Explain the split phase concept with an example 5 6 Explain the task function with an example 5 7 What is event driven programming and why is it critical for sensor network programming 5 8 Briefly discuss the following topics pertaining to nesC a Concurrency b Synchronous code c Asynchronous code 5 9 Specify a nesC interface for the following data types stack queue singly linked list and binary tree 5 10 Consider a data stucture composed of a stack and queue and assume that we want to build a datatype capable of manipulating the stack and queue individually Propose a nesC interface for this data structure REFERENCES 1 J Hill R Szewczyk A Woo S Hollar D Culler and K Pister System architecture directions for networked sensors In n Architectural Support for Programming Languages and Operating Systems 2000 pp 93 104 2 R Kumar and V Garg Modeling and Control Logical Discrete Event Systems Kluwer Academic Press 1995 3 P Levis and D Gay TinyOs Programmin
133. URE 14 2 Basic structure of the sensor node in our simulator structure node The different layers of the sensor node have gates to the other layers of the sensor node to form the sensor node stack A simple module with wireless channel function ality is used to communicate with these compound modules sensor nodes through multiple gates The functionalities provided by each module are described below with radio CPU and battery module forming the hardware model of the sensor node From To Upper Upper Layer Layer E t From CoOrdinator Layer Protocol To CoOrdinator L 1 From To Bottom Bottom Layer Layer FIGURE 14 3 Representation of a layer in sensor node www finebook ir SIMULATION DESIGN 259 14 4 1 Coordinator Module The coordinator formal proprietary name CoOrdinator module has the functional ities that coordinate the activities of the hardware and the software modules of the sensor node It is basically used for interlayer communication This module needs to be extended and functionality has to be added for access to properties of new hardware or consumers added As shown in the Fig 14 3 the CoOrdinator class has reference to all layers in the sensor node and all layers in the sensor node may access the CoOrdinator class implementation Thus through the coordinator module any layer may access and update the properties of the other layer For
134. V set is an independent set X where X vi not an independent set for any vertex v not in X The maximal independent set of largest cardinality is called the maximum independent set For the graph shown in Fig 4 46 some independent sets are shown below X i 1 4 8 10 X5 1110 6 3 X3 8 4 X4 1 7 10 Among these X and X gt are maximum independent sets All the vertices of an independent set can be colored with the same color Each vertex of a clique needs to be colored with a distinct color 4 14 INTERSECTION GRAPH A number of data structures graphs and trees that have been discussed previously must also address a different need integrating distributed data structures over sensor networks However for many deeply embedded systems we need a standard definition for a universal data structure that integrates various sensor network components efficiently into a single address space That is the topic in the next section of the chapter Let F be a family of subsets From the family F we construct a graph G V E as follows The vertex set V has one one one to one correspondence with F FIGURE 4 46 A graph with clique covering 1 2 3 4 5 6 7 8 9 10 www finebook ir 86 DATA STRUCTURES FOR SENSOR COMPUTING Two vertices of V are adjacent if and only if the corresponding subsets in F have a nonempty intersection The graph G constructed P in this manner from is called the intersection graph of F Let F P Po P
135. Z Mica2 Mica2Dot Flash memory kB 128 128 128 Measurement memory kB 512 512 512 EEPROM kB 4 4 4 A D channels 10 bits 8 10 bits 8 10 bits 8 Frequency MHz 1400 2483 5 433 868 916 433 868 916 Data rate kbps 250 19 2 19 2 Outdoor range m 100 300 300 Size 6x3x1 cm 6x3x1 cm 2 5x0 6 cm www finebook ir 30 SENSOR TECHNOLOGY FIGURE 3 5 MTS300 data aquisition board and a general purpose interface on the Mica2Dot motes Figure 3 5 shows a typical sensor data acquisition board Some other acquisition boards include the MTS300 MTS310 Fig 3 6 MTS101 and MDAS500 3 1 2 The Telos Tmote Sky Family The Telos family of sensors consists of TelosA and TelosB motes 5 They are a newer generation of motes when compared with the Mica family as they have a universal serial bus USB interface for data collection and programming In a similar design the Tmote sky sensors contain a USB port to facilitate pro gramming and are an exact replica of the TelosB suite of sensors These features make them well suited for wireless sensor network experimentation in the research community Figures 3 7 and 3 8 show the TelosB and Tmote sky sensors FIGURE 3 6 MTS310 data aquisition board www finebook ir SENSOR LEVEL 31 user button _ reset button TSR photodiode amp PAR photodiode LEDs SHT11 humidity temp 6 pin expansion USB serial 10 pin expansion reset support 1 bottom TI MSP430 F1611 S
136. ace Definition The following program shows the wiring called CollisionAvoidance tokenBased We need to build our component boot for booting the component LowPowerListening for minimizing power consumption Timer0 to run a clock with millisecond precision send to send packets of messages AMSend to send packets to intended destinations Receive to receive packets from other nodes and finally PacketAcknowledgement to indicate whether the packet sent is to be acknowledged For these interfaces we choose the following COLLISION AVOIDANCE TOKEN BASED APPROACH implementations as they are already provided in the TinyOS system configuration transmitAppC U implementation components MainC components Collision as AppC components CC2420ActiveMessageC components AMSenderC components new AMReceiverC components Timer lt TMilli gt as Timer0 AppC Boot MainC AppC Packet gt AMSenderC AppC Send AMSenderC Wrong component AppC PackageAcknowledgement gt AMSenderC AppC Receive gt AMReceiverC AppC Timer0 Timer0 AppC LowPowerListening gt CC2420ActiveMessageC Pseudocode Implementation event void Boot booted Start one shot timer of Timer0 using TimerO startOneShot t1 where t1 is sometime point in future when the clock fires once event void TimerO fired configure the LowPower Listening cycle of mote event void Receive received configure the LowPower Listening cy
137. ag 4 getdata line 2 file sensor visualization 1 getdata line 2 file 0 NOVO event type case 02 check if bag is over weight if str2num getdata line 4 file weightthreshold disp Bag getdata line 2 file overweight sensor visualization 1 getdata line 2 file str2num getdata line 4 file overweight event type pass bag overweight event to visualization else disp Bag getdata line 2 file weighed in at getdata line 4 file lbs sensor visualization 1 getdata line 2 file str2num getdata line 4 file ok event type disp Bag getdata line 2 file scanned error getdata line 4 file www finebook ir SOURCE CODE 289 sensor visualization 8 getdata line 2 file 0 getdata line 4 file event type time_introduced_to_bhs time random default have not handled o case where bag was never checked in for x 1 line if strcmp getdata x 2 file getdata line 2 file amp amp strcmp getdata x 3 file 02 tmp datevec getdata x 1 file time introduced to bhs datenum tmp2 tmp 4 6 1 end end times waiting for scan times waiting for scan time time introduced to bhs bags that were rescanned are currently averaged in twice possibly give 2nd scan different event id otherwise need more logic case 04 check if bag went in out of storage results 0 for x 1 line if strcmp getdata x 2
138. ailable memory The ease of modifying the sensor network and scalability which is defined as the number of nodes that can be simulated are two distinguishing features of this simulator 14 4 INTRODUCTION Wireless sensor networks WSN 5 consist of numerous tiny sensors that are de ployed in spatially distributed terrain These sensors are endowed with a small amount of computing and communication capability and can be deployed in ways that wired sensor systems cannot For example sensors can be deployed in environments that are in accessible for humans or sensor networks can be deployed in environments that are changing such as a chemical cloud Despite the prolific conceptualization of sensor networks as being useful for large scale military applications the reality is that the best migration path for sensor networks research into nonacademic applications is via integration with existing engineering application infrastructure For example sensor networks have the potential to offer fresh solutions to fault diagnosis health monitoring and innovative human machine interaction paradigms 4 20 13 19 Before emerging technologies such as sensor networks and the underlying node level architectures such as the event driven architecture of Tiny OS 11 can be incorporated as subsystems in mainstream engineering applications it is necessary to demonstrate the efficiency and robustness of these subsystems through compre hensive simulations that i
139. al 57 Heterogeneity 57 Heterogeneous Sources 58 203 Homeomorphism 58 Imote2 59 Implicit Embedding 59 Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 313 www finebook ir 314 INDEX Induced Subgraph 59 Installing TinyOS in Linux 60 Installing TinyOS inWindows 61 Intelligent Sensor 61 66 Intersection Graph 62 Isomorphism 62 Isomorphism Applications In Sensor Deployment 62 Kuratowski s Theorem 62 Link Layer Protocols 62 Linked List 63 Maximal Clique 63 Maximal Outer Planar MOP 63 Medication Device Protocol 63 Medium Access Control Layer 64 Memory 65 Mica Family 65 Microcontroller 65 Microelectromechanical Systems MEMS 65 Microframework 65 Minimum Cost Spanning Tree 67 Minimum Separator 67 Neighboring Nodes 68 NesC Programming 70 Network Layer 70 Network Localization 70 n Hop Neighborhood Group 72 Object Naming Storage Service 72 Operating System 75 Outer Planar Graph 76 Overhearing 77 Parallel Algorithm 78 Path 78 Perkins s Solution 79 Physical Layer 79 Planar Graphs 81 Planarity Testing 83 Power Unit 83 Protocol Programming 83 Publish Subscribe Group Pattern 84 Quality Of Service 85 Queue 86 185 Reduced Function Device 86 Regular Graph 92 Resource Constrain
140. al stimulus such as heat light sound or pressure and produces a measurable electrical Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc www finebook ir 4 INTRODUCTION Data Sinks q Sensing Nodes FIGURE 1 1 Networking structure of a distributed sensor network signal Thus a sensor with its own sensing device a memory and a processor can typically be programmed with a high level programming language such as CorJava The sensing devices can range from nanosensors to micro and megasensors In the remainder of this book when we refer to a sensor we refer to a whole system such as a mote which may have more than one physical sensor its memory processor and other associated circuitry Figure 1 1 shows a distributed sensor architecture and various components 1 1 2 Sensor Networks A distributed sensor network DSN is a collection of sensors distributed logically or geographically over an environment in order to collect data Distributed computing and distributed problem solving are commonly used in DSN in order to abstract relevant information from the data gathered and derive appropriate inferences This kind of data fusion can be used to compensate for the shortcomings of the in dividual sensor in real world enviornments For more details on sensor networks see Refs 1
141. algorithms These make use of lower band width for control packets resulting in less need to food the network for periodic updates This in turn results in energy savings for sensor nodes extending the life of a wireless sensor network considerably Examples of dynamic routing in wireless sensor networks include caching and multipath CHAMP routing and hierarchical state routing HSR We have now examined the structural characteristics of sensors and some of the unique properties of wireless sensor networks In the following section we will discuss the components forming the network stackin wireless sensor networks 83 SENSOR NETWORK STACK Communication routing and data transfer in sensor networks is possible between dif fering node types because of well established standards and specifications providing low level implementation details on how data can be exchanged in sensor networks www finebook ir 136 ALGORITHMS FOR WIRELESS SENSOR NETWORKS Application Layer Services Objects Vendor Defined Area v Network Layer Medium Access Control IEEE 802 15 4 Physical Layers FIGURE 8 4 Sensor network stack architecture The IEEE 802 15 4 is the standard specifying details on data exchange in the physi cal and medium access control MAC layers for low rate wireless networks When sensor network based applications are written most interactions with the IEEE
142. alifornia Berkeley 2004 9 G S Fishman Principles of Discrete Event Simulation Wiley 1978 www finebook ir 276 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION LSU Research Group LSU SensorSimulator LSU Sensim version 1 Jan 2005 User Manual Dept Computer Science Louisiana State University Baton Rouge J Hill R Szewczyk A Woo S Hollar D Culler and K Pister System architecture directions for networked sensors ACM Sigplan Notices 35 93 104 2000 C Intanagonwiwat R Govindan D Estrin J Heidemann and F Silva Directed diffusion for wireless sensor networking EEE ACM Trans Networking 11 1 216 Feb 2003 S S Iyengar and R R Brooks eds Distributed Sensor Networks CRC Press Dec 2004 D B Johnson The Rice University Monarch Project Technical Report Rice Univ 2004 C Mallanda SensorSimulator A Simulation Framework for Sensor Networks Master s thesis Dept Computer Science Louisiana State University Baton Rouge J Misra Distributed discrete event simulation ACM Comput Surveys 18 1 39 65 March 1986 OPNET Technologies Inc Opnet Modeler www opnet com S Park A Savvides and M B Srivastava Sensorsim A simulation framework for sensor networks Proc 3rd ACM Int DRAFT Workshop on Modeling Analysis and Simulation of Wireless and Mobile Systems 2000
143. alysis of events most tasks fall into these two categories such as node localization and target tracking Several methods exist to synchronize nodes in a network the most basic of which is sender receiver or receiver receiver synchronization In this approach peers in a network conduct time synchronization using timestamps in data packets as a reference A visual representation of the algorithm and the accompanying equations are shown in Fig 8 5 Some drawbacks exist with this basic method of synchronization Most notable are the four different types of delays involved Send time time spent in building the packet for transmission and delays inher ent in the protocol stack Access time time spent waiting for the medium to become free www finebook ir SYNCHRONIZATION IN WIRELESS SENSOR NETWORKS 139 Propagation time time taken for the packet to reach the destination Receive time time spent by the receiver processing the packet before it is timestamped Although send and access time delays can be eliminated with the use of reference broadcast synchronization some newer approaches exist that are less error prone We will now discuss the programming challenges involved in implementing a few basic algorithms 4 from the WSN area First the beaconing behavior of surrounding nodes is examined 8 4 1 Beaconing Beaconing refers to the continuous transmission of small control packets that notify neighboring nodes about
144. an Institute of Science in 1979 He then joined the Department of Aerospace Engineering as an assistant professor He is currently the associate director of the Indian Institute of Science and a Professor at the Department of Aerospace Engineering and at the Supercomputer Education and Research Centre He played a crucial role in building India s first Supercomputer Centre and the National Centre for Science Information at the Indian Institute of Science His areas of research in which he has published over 200 papers in international journals and for presentation at international conferences include numerical electro magnetics high performance computing and networks polarimetric radars and aerospace electronic systems information security digital libraries and speech www finebook ir ABOUT THE AUTHORS xxiii processing He has received many awards including the Padmashree Homi J Bhabha Award the JC Bose National Fellowship the Alumni Award for Excellence in Re search for Science amp Engineering at the Institute the Millennium Medal of the Indian National Science Congress in 2000 Ph D Honoris Causa from Punjab Technical University in 2003 and the CDAC ACS Foundation Lecture Award He was the NRC Senior Resident Research Associate at the National Severe Storms Laboratory Norman Oklahoma USA from 1987 to 1989 He was a visiting research scientist at the University of Oklahoma in 1990 Colorado State University in 1991 and has
145. and T of the same graph G the value of one spanning tree can be obtained from those of the other one by a finite number of cycle interchanges The number of cycle interchanges needed to obtain one spanning tree value from those the other of one is equal to the number of edges in one spanning tree that are not in the other We define d Tj T2 number of cycle interchanges needed to get T gt from T number of edges in T that are not in T3 3 I 2 5 4 9 I5 I 7 14 10 13 12 8 6 FIGURE 4 32 A spanning forest of G www finebook ir GRAPH TRAVERSAL 75 Lp a gt du J vl Lo FIGURE 4 33 a A graph G b some spanning trees of G c some more spanning trees of G We can also observe that d T T gt satisfies the following properties for a metric 1 d J T5 gt Q and d T T5 0 if and only if T T3 2 d T T5 d T gt To 3 d T T5 lt d T T3 d T3 T5 for any spanning tree T3 other than Tx or T gt Also d T T2 cannot exceed the rank or nullity of the graph Let G be a connected graph and T be the collection of all the spanning trees of G A graph can be formulate dusing J as a vertex set Let the graph be denoted by Gr T Er where two vertices Ti and T are joined by an edge and only if d 7 T2 11 The graph Gr is called the tree graph of G 49 GRAPH TRAVERSAL In order to visit all the vertices of a graph there must be some systematic way to ensure that no vert
146. arallel edge and no self edge is called a simple graph If the vertex set is an infinite set then the graph is called an infinite graph Otherwise the graph is said to be a finite graph A graph may have its edge set to be empty Such a graph is www finebook ir 62 DATA STRUCTURES FOR SENSOR COMPUTING a d ef FIGURE 4 15 A graph called a null graph A vertex in which no edge incidents is called an isolated vertex The number of edges incident on a vertex v is called the degree of v In the graph represented in Fig 4 15 f is an isolated vertex The degrees of the vertices are given below A vertex whose degree is 1 is called a pendant vertex In our Vertexabcde f Degree233310 graph e is a pendant vertex We usually denote the number of vertices and the number of edges by n and m respectively We are interested in the sum of the degrees of all the vertices Each edge incidents on two vertices so the presence of each edge contributes 2 to sum of the degrees Hence we have the following observations 1 The sum of the degrees of all the vertices is 2m 2 The number of vertices of odd degree is alway seven 4 7 2 Regular and Complete Graphs A graph in which all the vertices are of equal degree is called a regular graph If a simple graph has n vertices at the most n n 1 2 edges are possible A graph with all possible edges is called a complete graph Let K denote a complete graph with n vertices and n n 1 2 edges Note
147. are used to predict the received signal power of each packet These models affect the communicating region between any two nodes and are derived by the wireless channel 1 Free Space Propagation Model The free space propagation model assumes the ideal propagation condition that there is only one clear line of sight path between the transmitter and the receiver H T The received signal power in free space at distance d from the transmitter is estimated as 2 _ P G G X o0 Amd L www finebook ir IMPLEMENTATION DETAILS 261 where P is the transmitted signal power P is the received signal power G G are the antenna gains of the transmitter and the receiver respectively L is the system loss and is the wavelength 2 Two Ray Ground Reflection Model A single line of sight path between two mobile nodes is seldom the only means of propagation The two ray ground reflection model considers both the direct path and aground reflection path This model gives more accurate prediction at a long distance than the free space model The received power at distance d is predicted by _ P G G h x h B dix L P where h and h are heights of transmit and receive antennas respectively This last equation shows a power loss more rapid than that for the free space model as distance increases 14 4 4 Sensor Node Stack The simple module at the highest level of the hierarchy of the sensor node namely AppLayerSi
148. ation when running unattended A A node in GlomoSIM ON Application Mica2 P 3 Transport a Network MAC gt lt Y Physical Sensor 4 N l Battery lt P FIGURE 13 11 Simulator framework www finebook ir 246 PERFORMANCE ANALYSIS OF POWER AWARE ALGORITHMS TABLE 13 1 Data Gathering Reliability for Large Sensor Network Sensing Reliability Acknowledgments ARQ RTS CTS Priority Congestion control Fragmentation Link quality estimation 13 2 2 Communication Transparency The basic internode intercluster communication paradigm needs to be very efficient in terms of network latency and power savings and be able to adapt to traffic changes These parameters which are at the runtime framework level and independent of the application are listed in Table 13 2 A simple implementation uses leveled priority or standard queue to manage message pools As this subtopic is sensitive to node placement and harsh deployment conditions we assume a clusterlike topology of physical nodes that are balanced using a neighborhood table to schedule the message to forward clusters by minimizing communication overheads Refer to Figs 13 12b and 13 12c for multihop and single hop performance during application lifetime The service framework power savings optimization with deployment of node density in lifetime is shown in Fig 13 12d A service driven architecture approach defi
149. been a visiting professor at Carnegie Mellon University since 2000 He is an hon orary professor in the Jawaharlal Nehru Centre for Advanced Scientific Research JNCASR He is a fellow of the Academy of Sciences for the Developing World TWAS Indian National Science Academy currently the vice president Indian Academy of Sciences Indian National Academy of Engineering National Academy of Sciences and Institution of Electronics amp Telecommunication Engineers He is currently a member of the National Security Advisory Board part time member of the Telecom Regulatory Authority of India and member of the Board of Governors of IIT Chennai He is a directors of 1 Bharat Electronics Limited BEL 2 Data Security Council of India and 3 CDOT Alcatel Research Centre at Chennai he is a member of the Council of CDAC and the council of many universities and CSIR laboratories He is Editor of the International Journal on Distributed Sensor Networks He was a member of the Scientific Advisory Committee to the Cabinet SAC C a member of the Board of Governors IIT Delhi Chairman All India Board of Information Technology Education of AICTE and Editor of Electromagnetics and International Journal of Computational Science and Engineering until recently Chuka D Okoye Chuka D Okoye is an undergraduate student majoring in computer science in his senior year at Louisiana Tech University He has been named to the President s and Dean s lists mu
150. by its compiler linker For example a name server can provide the location transparently or it can be a remote procedure call generated by the compiler In addition the programmer must be able to debug and maintain the system by viewing and manip ulating the code in many sensors simultaneously Formal models and theory help in simplifying this complex task Because of the immense scale of DSNs techniques that support the automatic generation and analysis of software are important In an automated code generation system the responsibility for managing and maintain ing all the interactions between the sensors by message passing shared memory or sharing I O status is handled automatically Formal models and theory such as Petri nets or compiler transformation theory facilitate the task of software synthesis and integration by exploiting the underlying mathematical structure The user is responsible only for providing a high level specification of the application needs In addition the formal models and theory are also useful for introducing new function ality such as abnormal state recovery alarming and diagnostics The viewpoint is another important issue in the programming aspect Most of the current programming environments support a sensor centric view In this view the needs of the control ap plication must be expressed in terms of the capabilities of the sensor that is used in the DSN When dealing with large applications managing
151. c density nodes are subject to heavy congestion and jamming High bit error ratio low band width and asymmetric channels make communication unpredictable Because of the continual possibilty of loss of nodes and links due to power constraints dam age or eventual failure the connectivity and routing structures of the network will change dynamially This unpredictability represents many challenges for sensor net works including the design of communication protocols and development of data management techniques Heterogeneity Early research in wireless sensor networks focused on homogeneous architecture in which all sensor nodes possess identical software and hardware This architecture is resilient to individual failures More recently particularly in real world deployments heterogeneous sensor networks have become popular because of the advantage of increasing network lifetime and reliability Heterogeneity in sensor networks can be grouped into three categories computational heterogeneity where nodes have different computational power link heterogeneity where some nodes have long distance highly reliable communication links and energy heterogeneity where some nodes may have unlimited energy resource such as being connected to a wall outlet The network architecture of heterogeneous sensor networks likely has several tiers of nodes with different performance characteristics Heterogenous networks pose challenges to data processing techn
152. children of the root and they are assigned the level number 1 The root is called the parent of its children In our tree 1 and 3 are the children of 2 If a vertex v is at level i then any other adjacent vertex u that is not the parent of v is assigned level i 1 Such a node will be called a child of v If the maximum level number of a rooted tree is k then its height or depth is defined as k 1 The tree in Fig 4 21b is of height 5 A connected subgraph of a tree is called a subtree Figure 4 22 shows a tree and some of its subtrees A rooted tree is usually represented by its parent relation If T V E www finebook ir 68 DATA STRUCTURES FOR SENSOR COMPUTING 3 SubtreeTI 4 2 4 Subtree T2 10 5 E 4 8 FIGURE 4 22 A tree and someof its subtrees is a rooted tree with root r it is represented by the array parent 1 n where n is the number of vertices defined by PARENT i the parent of i if i is not the root PARENT r r where r is the root In some cases the parent of the root is defined as 1 A tree and its parent representation are shown in Fig 4 23 4 8 1 Binary Trees A rooted tree in which every vertex has at most two children is called a binary tree Binary trees are widely used in computer science applications The two children of a PARENT i 3 3 4 4 2 4 4 7 7 FIGURE 4 23 A root of a tree and its parent representation www finebook ir TREES 69 Left sk
153. cle of mote send acknowledgement to sender www finebook ir 146 ALGORITHMS FOR WIRELESS SENSOR NETWORKS extract token from packet transmit any data you need to send any node relinquish the token by sending to another node transmitAppC For low power listening we wire our components to the modules that implements the CC2420 radiochip functionalities For simplicity we provide the implementation of our component in pseudocode Transmit Application Configuration File When the monoshot timer triggers have fired we put our component into low power listening mode and then wait for someone to send us a token When we receive a token we send back an acknowledg ment perform any activities that we want to perform sending messages to any node etc and once finished we hand over the token to a neighboring node 8 5 2 Schedule Based Communication In order to minimize collision and save energy we need to follow a systematic way of transmitting and sensing behavior at each node In this section we write programs that follow a schedule so that the transmitter transmits only when the listener is listening and the listener is listening only when the transmitter is transmitting There are two ways a schedule can be embedded at each node explicit and implicit In the explicit embedding the schedule can be described using a data structure such as a list implemented using an array In the implicit embedding the schedule is implie
154. command error t send char msg 20 uint8 t len event void sendDone char msg 20 error t err PrintMessageC nc In terms of code this is a version of the HelloWorld program for sensor networks When the send command is called from another compo nent a message passed to the send command is printed on the screen using the built in debug function Note that this message can been seen only when run in a simulator since sensor nodes are devoid of any output except for some debug LEDs On printing the sendDone event is signaled to indicate completion of the command This is another illustration of the callback function discussed in the preceding chapter As stated before events are handled by the calling program in this case PrintClientC below which calls the PrintMessageC component The rationale for the implementation of events in the caller is be cause a module making use of an interface provided by another component is in a better position to deal with events generated due to the execution of a command Thus it can be said that PrintClientC makes use of the interface PrintMessage and a particular implementation of PrintMessage is provided by the component PrintMessageC module PrintMessageC implementation command error t PrintMessage send char msg 20 uint8 t len error t err if strlen msg lt 20 amp amp len lt 20 dbg STD Message was s n msg err SUCCESS Signal PrintMessage sendDone msg err
155. d in the sequence of operations that the node executes In this example we follow the implicit style We can build in the schedule implicitly at each node in its program as follows Let A 0 B mean at time t 0 A can transmit a packet to B Thus A 0 B B 10 C C 20 D D 30 niD E 40 A F 50 E can be a schedule Total period T 60 time units assuming that F 50 E takes 10 units of time The schedule will repeat periodically with a period T of 60 time units We further illustrate in this example how we can specify the behavior of the node abstractly using rules The value of the timer as specified by the time globally schedules the operations so that at each node s operations are executed as per the intended schedule A t 0 transmit data packet to B B t 10 transmit data packet to C C t 220 transmit data packet to D D NULL E t 40 transmit data packet to A F t 50 transmit data packet to E j www finebook ir COLLISION AVOIDANCE TOKEN BASED APPROACH 147 8 5 3 Pseudocode at Each Node Let t stand for the timer value We can implement this by building a counter that goes through states triggered by a clock module transmitC uses interface Timer lt TMilli gt as Timer0 uses interface Send uses interface Receive uses interface AMSend uses interface Receive configuration transmitAppC U implementation components MainC components transmitC as AppC components Act
156. d populateSensorTable myInfo id myInfo macAddress else if myInfo requestCode 0x0 Send my info to some other mote sendSensorInformation myInfo id call Leds lediToggle Toggle Led 2 to indicate request received return msg void sendSensorInformation uint16_t receiver message t msg Sensor sensorInfo call AMSend getPayload amp msg sensor id sensoraddress Sensor macAddress sensormac call AMSend send receiver amp msg sizeof sensorData www finebook ir SYNCHRONIZATION IN WIRELESS SENSOR NETWORKS 143 void populateSensorTable uint16 t id uint32 t mac 1 int i 0 Simple check if current mac already exists in table while i lt index amp amp table i macAddress mac i i if table i macAddress mac i table i moteID id else i table i moteID i d table i macAddress mac Implementation Details of sendSensorInformation and populateSen sorTable Additionally we need to provide implementations for the events in our program which are invoked when we call AMA setAddress AMSend send and RadioControl stop In sendDone when we fail to send a message we turn on an LED and then do nothing stopDone sendDone and Changed Implementations In wireless sensor net works concurrent transmission by multiple nodes could result in transmission errors due to packet corruption An example of such a scenario would be the widely stud
157. d G3 are said to be homeomorphic if one can be obtained from the other using a finite number of operations merging of series edges and or insertion of a vertex of degree 2 A complete subgraph is called a clique A clique that is not a subgraph of any other clique is called a maximal clique Consider the graph shown in Fig 4 20a Its maximal clique is shown in Fig 4 20b 4 7 6 Isomorphism Two graphs G V E and G V G are said to be isomorphic if there is a bijection between V and V in such away that two vertices of V are adjacent if and only if the corresponding vertices in V are adjacent If two graphs are isomorphic to each other then they have an equal number of 1 vertices 2 edges and 3 vertices with the given degree Note that these three conditions are necessary but not sufficient It is an interesting research problem to find a simple and efficient criterion to check whether two given graphs are isomorphic This problem is called the isomorphism problem www finebook ir 66 DATA STRUCTURES FOR SENSOR COMPUTING a 6 S02 2 3 5 I 2 P lt j 3 4 7 9 8 b FIGURE 4 20 a A graph G b maximal cliques of G 4 8 TREES A connected graph having no cycles is called a tree The following statements are equivalent 1 G is a tree 2 There is exactly one path between any two vertices of G 3 G is connected and contains n vertices and n 1 edges 4 G is minimally connected 5 G has no
158. d receiver common for mobile agents Hidden terminal problem a scenario in which the medium around the source node is free but busy around the destination node Exposed terminal problem in this case the medium around the destination node is free but engaged at the source node 8 3 2 Medium Access Control MAC Layer The MAC layer is the second lowest layer offering a management interface for the physical channel It is responsible for frame validation timeslot allocation synchro nization and node associations in a network There are several other implementations of the MAC Layer for sensor networks such as S MAC B MAC and C MAC all of which have different strengths ranging from energy conservation to routing speed gains References 8 6 and 11 provide more information on MAC implementations in wireless sensor networks 8 3 3 Network Layer Although there is no defined standard for the network layer in sensor networks several differing implementations exist today the most common of which is the Zigbee specification Details on the Zigbee network and application layer are covered in Chapter 10 8 3 4 Full Function Device FFD Full function devices are nodes having a general model of communication allowing them to talk with any other device Also such a device can be assigned the role of coordinator of a personal area network 8 3 5 Reduced Function Device RFD Reduced function devices are more restricted in their
159. d them with their requests to all their neighbors Obviously this way of passing data from a source node to a sink node is expensive but sometimes this may be the only way to send data In the following text we first present the pseudocode for flooding and then discuss the nesC programming challenges that we face while implementing it Node A the source node This node initiates the flooding from sensor or user Higher level behavior sense data packet p Data d lt packet p command forward gt build neighbourhood table table T for each node k in the t able T Or broadcast to every one send by unicast data d to node k sleep until woken up Node i where i is any node except node A build neighborhood table T wait for Data d Higher level behavior for each node k in the table T Or broadcast to every one send by unicast data d to node k sleep until woken up In the pseudocode above we have used two higher level behaviors in the imple mentation of our solution to flooding sense data packetp and sleep untilwoken up Both these behaviors are implemented using the S MAC schedule based listening and sleeping behaviors www finebook ir FLOODING 183 floodC lt Boo Initialize Read Receive Mainc FIGURE 9 3 Flooding protocol architecture interface initialize command initialize inte
160. dates need to be planned The items included in planning are tradeoffs of different updates relative to energy costs the injection strategy for network configuration and size reduction techniques that result in quick updates Injection strategies of software The strategies could include updating individual nodes or sending updates to a base station or to a number of select nodes that may then disseminate the updates to other nodes How software would be activated Software may be either automatically acti vated or based on a set of rules or manual activation may be required To meet the requirements for backward forward version compatibility control over the order of node activation may be needed Checking the downloaded software for integrity version mismatch and platform mismatch and dynamically checking the operation of the downloaded software after it has been activated Monitoring of update related faults Security related issues such as key distribution authentication secrecy in tegrity and authorization Problems related to very small nodes such as limited code memory and almost no RAM or EEPROM random access or electrically erasable programmable read only memory for storing new code Techniques may need to be developed for incremental building of new code into code memory usually flash RAM Version control that is prevention version mismatch Heterogeneity of sensor nodes There can be various forms
161. ddress Generate a new addr call Leds ledOOn A collision has occured generateAddress call AMA setAddress my group my address www finebook ir 176 TECHNIQUES FOR PROTOCOL PROGRAMMING broadcastAddress Only broadcast new address after a collision else call Leds led10n No collison yet return msg The monoshot time pulse triggers the event fired where we send the packet to the neighboring nodes This is done by constructing the payload and storing my address as part of the packet and then sending the packet After sending the packet if there are any errors reported in the send operation we turn an LED on After this we make no further attempts to send our address Whenever we receive a message from any node we extract the sender s address and compare it with ours to see if an address collision has occurred If no collision was detected then we have managed to find an address for our selves and remember this fact by setting the flag From this point onward we only need to respond to others messages by sending our address without changing our address This is achieved by setting the flag address Found to TRUE However if we noticed a collision we generate another address set it as our own address and broadcast it to all neighbors and then wait to receive their replies This continues until we find an address for ourselves Before we complete our task we need to add som
162. dling system CAN controller area network CBS checked baggage screening CHAMP caching and multipath routing DIFS distributed interframe space DSN distributed sensor network DVS dynamic voltage scaling EEPROM electrically erasable read only memory ESM electronic support measure FARM fusionable ambient renewable MAC FDMA TDMA frequency time division multiple access FFD full function device FIFO first in first out FSM finite state machine GEAR geographic and energy aware routing GHT geographic hashing table GPSR greedy perimeter stateless routing GSN Global Sensor Network GUI graphical user interface HAL hardware abstraction layer Proprietary organization abbreviations IEEE etc and very common acronyms e g CPU GPS IR PC UV omitted from this list www finebook ir xxvi NOTATIONS AND ABBREVIATIONS HBA HNG HPC HSR IFF INSPIRE IPC ISR LEACH LPL MAC MAS MCU MEMS MOP OMNeT PLC PSG QoS RAM ROM RFD RFID RSA RSS RTC RTOS RTS SCP SHIMMER SIFS SMS SPEED TCP IP TOSSIM UML USB UTC WASN WSN hub based architecture hop neighborhood group high performance computing hierarchical state routing IR infrared identification friend foe sensor innovation in sensor programming implementation for real time environment interprocess or intermediate performance communication intelligence surveillance reconnaissance low energy adaptive clustering hierarchy low power listening
163. driven topology of an actual sensor network using the idea of flooding INITIALLY i my own id r root node id neighbors Set of neighbors of i in the network parent NIL children empty set 0 others 0 START if i r and parent NIL then Send M to all neighbors parent i OnReceive Message M www finebook ir DEFINING DATA STRUCTURE OF SPANNING TREE PROTOCOLS 89 switch M gt type case NEIGHBORS if parent is NIL then Send parent to pj parent pj else Send reject to pj break case PARENT Add Pj to children if children UNION others contains all neighbors except parent then terminate break case REJECT Add Pj to others if children UNION others contains all neighbors except parent then terminate break 4 15 5 Model Approach We need to consider the case of synchronous versus asynchronous networks In a synchronous nework the fooding message always starts from the root and foods to the destination nodes The longest path on the network is also called the diameter D In a asynchronous network the fooding messages can be initiated simultaneously from each and every node and reach neighbors that are within range This is sometimes called on demand fooding where any node that needs to send data starts fooding to find its destination on the network 4 15 6 Complexity This allows us to compare the performance of both types of fooding protocol in terms of co
164. dth issues encountered in the first case In Fig 15 1 the sensors process the data and relay these intermediate results to the data fusion center which in turn processes the incoming information The global results are finally available at the data fusion center In energy and processing rich environment like an airport the first approach is appropriate The baggage is checked in at any of the desks labeled desk 1 desk 2 and desk 3 in Fig 15 2 During checkin RFID tags are attached to them After checkin they are placed on the respective conveyer belts and introduced into the BHS The scanner that is in close proximity to the location where the bags are introduced reads the RFID tags attached to the bag and energizes the corresponding diverter to feed it to the input of an X ray scanning machine If the X ray scanning machines are currently occupied the bag s RFID will be saved in a local queue and the bag will have to follow the loop until it is scanned again by the abovementioned scanner After the X ray machine the state of the bag will be any one of the following The bag passed X ray scan and deemed eligible for transportation to the aircraft The bag was identified for further manual inspection level 4 inspection The bag was not able to make its way through the X ray machine or it was not recorded The other scanner placed on the loop with the help of diverters dispatches bags for further inspection or diverts them to a cha
165. dule PacketSenderC continued event void AMSend sendDone message t msg error t error 1 if amp pkt msg amp amp error SUCCESS dbg x Packet was sent successfully n busy TRUE dbg x Packet was not sent n event message t Receive receive message_t msg void payload uint8 t len setLedBits 0x7 received data successfully dbg x Received a packet n return msg J PROBLEMS 7 1 What is the role of timers in wireless sensor applications 7 2 In every TinyOS based application what interface is responsible for system initialization 7 3 What are some of the challenges of writing wireless sensor applications www finebook ir 130 SENSOR PROGRAMMING 7 4 What role does the SplitControl interface play in TinyOS 7 5 List the differences between SplitControl and StdControl interfaces REFERENCES 1 Crossbow Telosb Datasheet Courtesy Crossbow Inc 2 J Hill R Szewczyk A Woo S Hollar D Culler and K Pister System architecture directions for networked sensors in n Architectural Support for Programming Languages and Operating Systems 2000 pp 93 104 3 P Levis and D Gay TinyOs Programming Cambridge Univ Press 2009 www finebook ir 8 Algorithms for Wireless Sensor Networks Talk is cheap Show me the code Linus Torvalds An algorithm can be defined as a logical sequence of instructions for solving a problem in a finite sequence of steps In wireless
166. e The data from the region follow the path established by the reinforced messages The nodes in the region send out data at the rate that is specified in the query Data caching is implemented in intermediate nodes and so the data requested by different subscribers from the same region can be satisfied by the common node in the path thus reducing the traffic and redundant messages The data marked as exploratory are sent to identify better paths and reinforce at regular intervals Also the neighbor updating procedure is carried out specifically at regular intervals the beacon messages are broadcast and beacon reply messages are sent by neighbors thus maintaining the latest neighbor information 14 5 2 802 11 MAC The MAC layer places the network packet on the wireless channel The network packet may be a broadcast or unicest packet to a specific node sink node Any www finebook ir 264 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION network layer packet received by the MAC 802 11 7 22 3 module is encapsulated into the MAC frame with the MAC header added to it The network layer packets have information on whether the packet has to be broadcast or unicast The broadcast packet is encapsulated into the broadcast MAC frame with the appropriate MAC header and is inserted into the messages queue of the MAC layer If the network packet is for a particular destination an RTS frame is created and is inserted in the messages queue of M
167. e X to learn more about the event The program below shows what the node should do depending on what its role is A node can play any of the three roles shown in the pseudocode presented below source role Program Code Node X Rumour initiator wants to report an event e to another node Z build neighbourhood table T for each node k from T d randomly decide to send if d yes send to node k own address event e end for loop Sleep until woken up Other intermediate nodes Y 1 flag 0 build neighbourhood table T if flag 1 display I was selected once before NT roe and I forwarded the packet www finebook ir 186 TECHNIQUES FOR PROTOCOL PROGRAMMING else for each node k from T 1 d randomly decide to send if d yes send packet to node k own address event e flag 1 read data and perform rumour routing sink role receive and process data intermediate node role if received first time flood the network Assume Let T be the neighborhood table int self first Time am_addr_t sensoraddress 2 Pre determined address of current sensor am group t sensorgroup 1 RoutingData table NO_OF_MOTES Assume we already have this table booted self initialize Role firstTime true call RadioCont rol start event void RadioControl startDone error_t err 1 Fi
168. e computing and the interac tion with the communication interfaces the execution in a sensor can be deterministic www finebook ir 46 DATA STRUCTURES FOR SENSOR COMPUTING quasideterministic or nondeterministic One of the main challenges in DSN research is to design efficient deterministic and quasideterministic sensor nodes 42 COMMUNICATION CAPABILITIES The communication subsystem is the primary infrastructure on which the DSN is constructed and hence design choices made in this subsystem strongly affect the other capabilities of the DSN Figure 4 3 presents a taxonomy of this subsystem The primary functions in this aspect are data transport and bridging We distin guish between three types of data each having different characteristics Input data gathered by sensors are typically limited to a few bytes and need guaranteed de terministic message delivery to maintain integrity Sensors communicate primarily to synchronize and to recover from failures Thus intersensor traffic is likely to be sporadic contain more information aggregated data and be more suitable to quasideterministic or nondeterministic delivery mechanisms System data refers to all the other data delivery needs that may or may not have hard real time require ments For example data required for system monitoring and status alarms may be critical and real time whereas data used by Internet based supervisory systems may not Non real time system data such as dow
169. e content based addressing in which codes are used to identify the type of data within the message Source or content based schemes are typically used on a broadcast bus a ring a data server or when routing schemes can be a priori specified Destination based schemes are used when there is usually one destination or when the routing is constructed dynamically The capability to provide deterministic service is strongly affected by the access method that establishes the rules for sharing the common communication medium www finebook ir 31o ourelJ suontgeorunululoo IOSUOSS pojnqrnsrp L jo AUIOUOX V t MNO pied C es SSeuppy CRIS C Suy C vna wopuey uao C Paica Y jse peoug jenas uejdopeg pare uuo Gag uonzunsqi jutog o14ulog sn e8pug S4 SuJUJO7 Josuas poy ss955v Suisseuppy Ck ule1sAS K8o odo je isAud Sulgexpeg e ejedu s d eq jo2010ud DYW peseg e unog Josues Jeu uBpug Croa uoneueuis dul uoun SHER ums Sumus summum Rande dw NSG 47 www finebook ir 48 DATA STRUCTURES FOR SENSOR COMPUTING Polled token based and time division multiple access TDMA schemes that use a fixed timeslot allocation are deterministic Token schemes which allow nodes to skip their timeslot when they have nothing to transmit have quasideterministic be havior Random access schemes such as Ethernet result in nondeterministic perfor mance and a
170. e g ni when sensing an object within its range raises an alarm by turning its red light ON and at the same time sends an alert message to the other node n2 which turns its green light ON We need to implement this in nesC Give the interface descriptions and show the implementation of each interface function Also show the final overall wiring diagram REFERENCE 1 P Levis and D Gay TinyOs Programming Cambridge Univ Press 2009 www finebook ir PART III Sensor Network Implementation www finebook ir 7 Sensor Programming Programming is an artform that fights back Chad Z Hower As sensing devices become more complex applications are being developed that exploit the increased power these devices provide Traditional programming models and abstractions only inadequately capture these newly found evices and thus there is a need to develop different types of techniques that naturally exploit the power of these sensing devices Sensor programming models should essentially support abstractions over collections of sensor devices heterogeneous data emanating from these devices and their data storage capabilities A sensor network can be viewed from several angles Our view is committed to a programmer s perspective where a programmer would like to program the nodes in the network at a fairly detailed level and at the same time is interested in programming the network at the abstract group cluster level At the node
171. e global addresss for each specific node and test if the address is in use FR k k k k k k k k k k k k k k k k k k k K k k k KOK K K K KOR R K R R R R K KOR R OR K R R R R K K S event void RadioCont rol startDone error_t err First we generate the address of the mote generateAddress The initial address broadcast call TimerO0 startOneShot 6000 At this point we can set the generated address call AMA setAddress my_group my address event void AMSend sendDone message t msg error t err if err FAIL call Leds ledoOn event void TimerO fired message t msg am addr t temp www finebook ir 172 TECHNIQUES FOR PROTOCOL PROGRAMMING temp am_addr_t call AMSend getPayload amp msg temp amp my address call AMSend send 0 x ffff amp msg sizeof am_addr_t RRR k k k k k k k k k k k RR RK k k RR K k k RR k k RR RRR RR RRR ke K R RO RK k ke ek A simple function that randomly generates an address for a mote RRR k k k k k k k k k k k k k k k k k k RR K k k K R K k KOR R R K k RO R R K k R R R K R R R R RK void generateAddress call Leds set 0 my address 31071334523 call Random rand16 void broadcastAddress message t msg am addr t temp temp am_addr_t call AMSend getPayload amp msg temp call AMA amAddress call AMSend send 0 x ffff amp msg sizeof amaddr t async event void AMA changed event
172. e is already implemented so it will suffice if we learn how to call its various commands and the event Given below is a simple application where we have used one command function and the event function from the interface Timer As we know by now an applica tion will consist of interface definitions their implementations using modules and configurations showing how the components are wired together In this example we use three interfaces Timer Leds and Boot and all these interfaces have already been implemented in the system so our only task is to implement only the events defined in these interfaces module Twinklec uses interface Timer lt TMilli gt as MyTimer uses interface Leds uses interface Boot implementation event void Boot booted call MyTimer startPeriodic 200 event void MyTimer fired call Leds ledOToggle This program describes a module Twinkle where we have shown an implementa tion for the event function booted from the interface Boot and the event function www finebook ir SENSING THE WORLD 119 clock pulses E FIGURE 7 2 An illustration of where Timer drives ledO fired from the interface Timer and have inserted calls to commands that toggle the red LED Let us now explain how the control flow occurs in this code To begin with note that we have chosen a time of millisecond precision and renamed it as MyTimer When we run this application the TinyOS system will beg
173. e location of the nodes in order to be implemented This pattern can also be used for local sensor selection Exceptions can occur when communications fail causing instability in hop counts Solutions to this problem can include fooding or ad hoc routing tree explicit or implicit www finebook ir PROBLEMS 205 10 3 5 Publish Subscribe Group PSG Pattern This pattern is useful in situations where nodes of shared interests come together to perform a common task For example this works when all agents that provide certain data work together in data fusion and forward the results to an application This pattern can be used in applications such as data service discovery and pursuer evader games However exceptional situations can arise when multiple nodes with shared interest food the network Protocol support such as directed diffusion which does not need geographic information GHT geographic hashing table hashing data into geographic locations and multirendezvous regions replication of GHT in a region may alleviate this problem 10 3 6 Acquaintance Group AG Pattern This pattern defines a group of nodes that used to know each other This captures the notion that all nodes that share some state with each other can fruitfully form a group This is particularly suitable for nodes that are mobile They may have peer to peer or leader follower structure and group membership can be historical or logical The challenge in appl
174. e more functions shown below void generateAddress call Leds set 0 my address 31071334523 call Random randl6 async event void AMA changed event void RadioControl stopDone error_t err We generate a function using a large random number to minimize address colli sions and quick and definite convergence We also note that we have nothing to do when ever the events AMA changed and RadioControl stopDone are invoked Note that whenever the command AMA changed is called AMA changed eventis signaled Similarly RadioControl stop sig nals RadioControl stopDone Where is RadioControl stop called www finebook ir IMPROVED ALGORITHMS 177 9 10 IMPROVED ALGORITHMS We will now discuss a couple of solutions that address the problems that we mentioned above 9 10 1 Perkin s Solution We will illustrate this solution with an example in which we are given a simple network with only a few nodes labeled To start with no nodes have any addresses assigned to them Consider node A which wants an address for itself It generates a pair of random numbers lt t f gt where t is the temporary address and f is the fixed address and floods the network with this pair Each node will receive a copy of this pair When for example node B receives this pair it checks whether if it already has an address that is the same as the number f Since node B does not have any address to itself as
175. e situations are both treated as bugs no different from running out of execution stack space with potential consequences that are just as disastrous Because of the crucial importance of data structures this entire chapter deals with various data structures used in the design and analysis For completeness all the data structures are discussed Those who are already familiar with the data structures may skip these topics An array usually represents a collection of homogeneous data items An array is also a list For example the messages of a given sensor node are usually represented in the form of an array as shown in Table 4 2 In the design and analysis of algorithms we frequently come across two special types of arrays stack and queue 4 4 1 Stack A stack is a one dimensional data array in which addition and deletion take place from one end Suppose that a stackcontains five data items Then a designated variable top 5 denotes that there are five items where item 5 is the topmost entity If anything TABLE 4 2 Stack Items Node ID 1 2 Received Messages 2 1 AU ON www finebook ir 52 DATA STRUCTURES FOR SENSOR COMPUTING lt Stack Top 4 Stack Rear FIGURE 4 5 A stack with five entities new has to be added to the stack it can be added as the sixth item and the top will now become item 6 From a stack with five entities see Fig 4 5 if we want to delete one entity we can rem
176. eal time continuous ordered sequence with limited control over the order in which items arrive and the limita tions of low battery life low bandwidth and low processing power and operating memory present programming challenges that are unique to the sensor network environment 13 SENSOR NETWORK SOFTWARE A network architecture and protocols are essential foundations for building software applications Developing computational communication systems for deployment and applica tion for wireless sensor networks has been a challenge since the mid 1990s More Specifically wireless ad hoc sensor networks have been largely designed with static and custom architectures for specific tasks thus providing inflexible operation and interaction capabilities WSN applications need to be programmed with constrained memory and process centric resource requirements in mind in order to write com munication code with real time sensing deadlines which are critical to a dedicated scheduled measuring task In short the problem is the choice of abstraction for the sensor node runtime environment Our computational framework or paradigm called INSPIRE defines and supports nanofootprint and real time deadlines sched uled tasks for computing and allows communication and sensing resources at the sensor nodes to be efficiently harnessed in high density event driven application sensing fashion through the use of an object oriented framework A key feature of the r
177. earch on the use of wireless sensor networks for forest fire detection has attracted considerable attention from the research community Firebug is a small wireless sensor for collecting real time data in forested areas developed by UC Berkeley Equipped with a position system device this sensor can collect data such as relative humidity temperature and pressure in situ and communicate with the remote data server through base stations Thus users can use the Internet to monitor the corresponding area The relative risk of forest fire danger with respect to sensor measurement data such as smoke windspeed and light level has been www finebook ir SENSOR NETWORK APPLICATIONS 19 investigated Flood detection is another application in environment observation An alert system is used to evaluate the possibility of potential fooding This system is equipped with wind temperature and water level sensors and is able to provide real time water level and rainfall information Wireless sensor networks also can be used for atmosphere pollution monitoring and detection of other natural disasters such as earthquakes and tsunamis 2 1 3 Health Applications Hospital and medical facilities can be improved by smart sensors For example patients with sensors on their bodies can be monitored and tracked inside a hospital or be talemonitored in their homes Medical applications have unique demands such as extreme robustness very dense networks and preservi
178. ecision of a clock that we want TMilli refers to a precision of one millisecond 1 kHz clock T32khz refers to a 32 kHz clock and TMicro refers to a microsecond clock Sometimes in our programs we will need a regular clock that when execution starts begins running as any real world clock would do Such clocks are im plemented using the startPeriodic command Suppose that we have an interface Timer lt TMilli gt s then startPeriodic 250 of this timer will start a 250 kHz clock that fires 250 x 1023 pulses every second starting from now until this timer is stopped where as startOneShot 250 fires only one pulse after 250 milliseconds starting from now and then stops An operation that is important to sensor programming is provided by the event fired Whenever a timer fires a pulse the TinyOS system invokes the event fired tosignify the fact that time has advanced by a unit of time We will be using this feature often whenever we use the Timer interface This also of course means that we need to implement the event function fired since we use the interface Sometimes it is useful to start the timer after some delay and we can do this using the command for example start PeriodicAt 5 250 which starts a periodic timer with a precision of 250 milli seconds beginning 5 milli seconds from now Other commands are described in the system and we will illustrate how to use some of them in the examples below Fortunately the Timer interfac
179. ect data periodically 4 Distributed algorithms for energy and reusability loading and fault tolerance in large sensor networks www finebook ir 6 INTRODUCTION 1 2 3 Wireless Sensor Network Environment Sensor Network make it possible to monitor instruct or control various domains such as homes buildings warzones cities and forests Sensor networks can ob serve the sensing environment at a close range and thus have many advantages such as ability to monitor smallest details proximity to places which are difficult to reach by humans for example difficult terrain or hazardous environment The major limitations of sensors are their limited power supply limited communica tion bandwidth and range and limited computation ability and memory capacity Data transmission consumes a large percentage of energy reducing the amount of data transmitted is the primary focus of data processing The small bandwidth of the wireless links represents a challenge for data processing Because of the lim ited communication radius of a sensor node data may have to go through multiple hops to reach the final destination This leads to extra power consumption in sen sor nodes on the relay path Limited processing and memory capacities restrict the complexity of data processing algorithms running at the sensor nodes The inter mediate results and other data are also burdensome to store in the node because of limited memory size Sensor data are a stream a r
180. ected graph and let S be a subset of V If G S is disconnected S is called a separator of G A separator S is called a minimal separator if whenever S is a subset of S S is not a separator A minimal separator of minimum cardinality is called a minimum separator The cardinality of a minimum separator is defined as the vertex connectivity of the graph In other words the vertex connectivity of a connected graph is the minimum number of vertices to be removed from the graph in order to make the remaining graph disconnected Note that if a www finebook ir 78 DATA STRUCTURES FOR SENSOR COMPUTING I 2 2 4 7 3 5 6 a l 2 8 7 3 5 6 b 2 8 3 4 7 5 6 c FIGURE 4 36 a Graph G b graph G 4 c graph G 1 graph has an articulation point the vertex connectivity of the graph is 1 A graph with vertex connectivity is called a separable graph A nonseparable graph is also called a biconnected graph For a biconnected graph the vertex connectivity is at least 2 A connected graph with vertex connectivity of at least 3 is called a triconnected graph In other words a triconnected graph is a connected graph in which the removal of 7 10 FIGURE 4 37 A graph www finebook ir CONNECTIVITY 79 FIGURE 4 38 A triconnected graph any one or two vertices does not make it disconnected Observe that any triconnected graph is biconnected and a biconnected graph is connected A connected graph with vertex connec
181. ed Computing Environment 93 99 Reusability Index 95 Rooted Tree 96 Routing 105 RTOS Abstraction Layer 107 RTS CTS Handshake 108 Rumor Routing 108 Schedule Based Communication 113 Security 134 Sensor computing 134 Sensor Level 135 Sensor Network Applications 136 Sensor Network Software 136 Sensor Network Stack 136 Sensor Networks 136 Sensor Node 138 Sensor Programming 140 Sensor Technology 140 Sensors 140 Separable Graph 141 Series Edges 142 Server Level 147 Shimmer 150 Shimmer Properties 150 Shortest Spanning Tree 150 Sleep Sheduling 153 Spanning Trees 159 Split Phase 159 Stack 159 Streaming 160 Subgraph 162 Subtree 162 Telos Tmote Sky Family 165 Time Synchronization 165 Token Based Approach 165 Tracking 167 Transceiver 181 www finebook ir INDEX 315 Transmitter Role 183 Wireless Channel Model 212 Transport Layer 188 Wireless Sensor Network Topologies 214 Trees 192 Wireless Sensor Networks WSN 221 Wiring 230 Uncertainty 209 WSN Protocol Stack 234 Unpredictability 209 ZigBee Application Development 252 Walk 209 ZigBee Devices 264 www finebook ir
182. ed in an ad hoc structure to form distributed sensor networks for sensor processing in a distributed manner Such net works have greater fault tolerance and sensing accuracy and are less expensive than are traditional networks Another interesting property of wireless sensor networks is that nodes can be deployed in hostile environments to provide continuous monitoring and processing capabilities for a broad variety of applications More importantly a sensor node integrates hardware and software for sensing data processing and communication For more details on sensor integration see an earlier text by the author 3 Furthermore these nodes can be deployed in large numbers in unstruc tured environments the nodes in these networks rely on wireless channels for the transmission and receipt of data from other nodes Communication among nodes can be characterized by the type of sensors that make them up for example RF sensors used in Berkeley motes have a maximum operation range of around 100 ft One interesting property of wireless sensor networks is a parameter termed sensing area Which depends significantly on the physical types of sensors being used For example a range sensor such as a range polarized 6500 ultrasonic ranging module used in many robotics applications can detect a target from as close as 6 inches in CHARACTERISTICS OF SENSOR NETWORKS 21 away up to a maximum distance of 35 ft Figure 2 2 shows an example of a distributed
183. ed in the Read module of the System When the data is read by the read command readDone event is invoked which calls the command leldOon to turn the light on Courtesy of TinyOS authors Kevin Klues and Joe Polastre description TinyOS StdControl Interface interface StdControl command error_t start command error_t stop www finebook ir 122 SENSOR PROGRAMMING invoke event call command TinyOS system Boot booted invoke event Timer StartPeriodic Timer fired call command call command invoke event Read read Read readDone NENNEN Leds ledOon FIGURE 7 3 Control flow shown using an AND OR graph 7 33 APPLICATIONS USING THE INTERFACE SplitControl SplitControl is an interesting interface that provides an opportunity for the program mer to write programs in an embedded system where a physical hardware system and its equivalent software simulated system can be used interchangeably Accordingly SplitControl can also be viewed as an extended concept of StdControl and as the split phase counterpart to the StdContol interface The TinyOS programming manual 3 delves into more details Courtesy of TinyOS authors Kevin Klues and Joe Polastre description TinyOS SplitControl Interface interface SplitControl command error t start event void star tDone error t error command error_t stop event void stopDone error_t error event void stopDone error_t e
184. ed systems 1 data or program location 2 data or process replication 3 process migration 4 concurrency and 5 parallelism For our purposes in a www finebook ir X10 oure1 Surnduroo Josuss pojnquusip e jo AWOUOXeL cr AWAD peziuouu uAS Aen S e S J91se A aw APD Ga CN Te GD Buuu PaO sweJSoJy SJOLUSY CEED M JeA JeSuel eye 9 91 uls 98eJo1s uoneziuo4upuAS M Gurea 9 9 HIN SODIA I9S 9J4nj 91lu 4 peinqiusiq 20525204 ulo1s4S Sune4ed Josu s Se ejJo1u WSS sonsougel sse oug jueuuedeue q Suisse oJg uuog y uonejueuueduul uonpunj seinquay ui ls 4s SuiuueJSo4g suoneoiunuJulo Sunnduuo mdu NSC 44 www finebook ir INTRODUCTION TO SENSOR COMPUTING 45 DSN transparency concerns the object naming storage service which provides the ability to access system objects without regard to their physical location and remote program services which provide the ability to create place execute or delete a program without regard to the sensor Typically servers are necessary to perform the registration and lookup functions to provide these services The atomicity service is used to increase the reliability of the system by ensuring that certain operations called transactions occur in their entirety or not at all Various forms of recov ery mechanism can be implemented to checkpoint and restore the component
185. ed to control the three LEDs Since the module PowerupC see www finebook ir PROGRAMMING CHALLENGES IN WIRELESS SENSOR NETWORKS 115 PowerupAppC PowerupC Qo Provides O Uses FIGURE 7 1 Component architecture for PowerupAppC also Fig 7 1 uses the interfaces Boot and Leds we need to provide implementation for both of them Courtesy of TinyOS authors description interface Leds Philip Levis and Joe Polastre TinyOS Interface Leds Turn on and off individual LEDs async async async async async async async async async command command command command command command command command command void void void void void void void void void ledoOn edooff ledOToggle ledi1On ed1Off lediToggle led20n led2Off led2Toggle Get the mask bits for the LEDs async command uint8 t get Set the mask bits for the LEDs async command void set uint8_t val In order to provide implementation for these interfaces we browse through the modules and configurations provided in the TinyOS system and we find that the configuration MainC provides an implementation for the interface Boot and the configuration LedsC provides an implementation for the interface Leds We specify this implementation of the interfaces by the following wiring statements PowerUpC Boot MainC PowerUpC Leds LedsC can also be written as LedsC
186. ed with the Microsoft net microframework preinstalled Figure 3 10 and Table 3 3 provide more information about the functional and physical characteristics of the sensor platform 3 1 4 SHIMMER SHIMMER which represents sensing health with intelligence modularity mobility and experimental reusability is a sensor platform for health related technologies It FIGURE 3 9 A TelosB sensor attached to the bumblebee radar board 7 www finebook ir SERVER LEVEL 33 FIGURE 3 10 An Imote2 sensor supports wearable applications such as capture of real time kinematic motion and physiological sensing SHIMMER motes are driven by TinyOS and support up to 2 GB gigabytes of data storage for offline data capture microSD storage Some applications of this platform include sleep studies cognitive awareness vital signs monitoring and chronic disease management see Table 3 4 and Fig 3 11 Apart from the few platforms briefly covered in this chapter there exist several equally important sensor platforms not covered in this book such as CSIRO s fleck platform for environmental monitoring SNoW5 platform and many others 3 2 SERVER LEVEL Gateways and programming boards that handle the buffering of data from the wire less network constitutes most devices at this level Most programming boards are dual purpose devices allowing direct access to motes for in system programming and at the same time serve as gateways for communication with an
187. efers to a sensing device that belongs to a wireless sensor network that is capable of processing gathering and communicating sensory information with other devices in the network The architecture of a typical mote is shown in Fig 8 1 With advances in microelectromechanical systems MEMSs and low power wire less technology the major components of sensors as shown in Fig 8 1 have been miniaturized over the years However the processing and storage capabilities of sen sors have not developed in accordance with Moore s law In the following section we compare the properties of some of these components microcontroller transceiver memory power unit and sensors with their traditional counterparts Power Generator FIGURE 8 1 Structure of sensor nodes www finebook ir STRUCTURAL CHARACTERISTICS OF SENSOR NODES 133 8 1 1 Microcontroller The microcontroller is one of the most important components in a sensor It is respon sible for data processing memory management interrupt handling and the control of other components on the sensor to minimize energy usage and increase the lifespan of the motes Unlike the microcontrollers present in traditional computers most sensors operate on less than a watt of power giving them a lifespan of several months to a year This power conservation is achievable because of the ability of microcontrollers to enter low power states when idling These microcontrollers have frequenci
188. eive check if all acks are received and stay in this state and if a 1 l are ve assert f as own address if any is ve call generate if it is a request for verifying some other node s request then extract f t part and check if fit clashes with its own address if it does send ve ack to t if it doesn t send ve ack to t 9 11 CONTENT BASED ADDRESSING Often in applications a node may be interested in knowing about events that other nodes may have detected It is possible that more than one node may have detected the nodes events and stored the data in their memory We now discuss a simple protocol where a node can query regarding an event to an unknown node The node that is interested in the event called the sink node generates a query and this is passed on to all nodes in the network The node that can produce an answer to the query called the source responds to this by generating data packets and forwards them to the sink The intermediate node in addition to forwarding the data to the sink node also stores the data along with the query in its cache which it use later for any similar queries from future sink nodes Consider for example the network shown in Fig 9 2 where node j green has detected an event E and node c yellow is interested in knowing about this event but it does not know about node j Node c then sends a query along with its address which is passed on to other nodes in the network
189. em starts executing the event booted which is where our program execution begins as far our program is concerned Once the execution of booted is started we call PrintMessage send which prints the message Hello World After printing this command signals back to invoke the event PrintMessage sendDone to indicate whether the command was executed correctly and prints a message appropriate for the error reported by the command We can now show the overall structure of our program rather component using the component diagram in Fig 6 1 configuration PrintClientAppC implementation components MainC components PrintMessageC components PrintClientC PrintClientC Boot MainC PrintClientC PrintMessage PrintMessageC www finebook ir A SIMPLE PROGRAM 103 PrintClientAppC Init MainC PrintClientC e e Init Boot Boot init booted PrintMessage PrintMessageC PrintMessage send sendDone FIGURE 6 1 Component architecture for PrintClientAppC Although this program is a simple one it illustrates several issues that one has to deal with while writing a nesC program which can be summarized as follows Define an interface I Provide implementation for the commands from the interface in a module M Provide implementation for the event from the interface in a module M that invokes the commands Provide a configuration module show
190. en done using a program ming stack model and here in this architecture as compared in Fig 13 3 a multihop FARMS Application Event Processor Object FrameWork RT Kernel Target FIGURE 13 4 Stack architecture www finebook ir SERVICE ARCHITECTURE 243 ypesssssesssssssessessscssesssssssensessesssscccessssscoconeeees 2 sssssssssssssssssssssssssos Network Network Service Protocol 3 Manager Neighbor Table Msg Pool SP FARMS Adaptor B Data Link B FARMS Adaptor A Data Link A JO3euns3 yur Jozewns3 ur FIGURE 13 5 Service architecture data forwarding service is emphasized Also from the traditional distributed model which calculates lifetime using fixed resources we extend the constrained resources by providing a dynamic renewable energy resources model We essentially refer to this process as adapting to the ground truth of an ambient wireless phenomenon In addition the stack model supports the Rx idle recharge cycle and energy wastage Message Message Transmission Reception Message Message Transmission Reception FIGURE 13 6 Traditional opaque layering in stack model www finebook ir 244 PERFORMANCE ANALYSIS OF POWER AWARE ALGORITHMS Message Message Transmission Reception A i SP l I I m o 1J ie i ES 19 Q I I i I I I v i Message Message Transmission Reception FIGUR
191. ensor network 13 2 List three ways in which the factors explained above can be improved to extend the life of a sensor network 13 3 List some advantages that the INSPIRE framework provides to existing sensor networks 13 4 How much of a performance hit is taken when power aware routing protocols are employed in sensor networks 13 5 Compare two 2 other network simulators with GlomoSIM discussed in this chapter REFERENCES 1 V Iyer S S Iyengar N Balakrishnan V Phoha and M B Srinivas Farms Fusionable ambient renewable MACs Proc IEEE Sensors Applications Symp SAS 2009 Feb 17 19 2009 pp 169 174 2 V Iyer S S Iyengar G Rama Murthy M B Srinivas and B Hochet Multi hop scheduling and local data link aggregation dependent qos in modeling and simulation of power aware wireless sensor networks ACM IWCMC June 21 24 2009 Leipzig Germany 3 V Iyer G Rama Murthy M B Srinivas and B Hochet C error simulator for development for sensor and location aware sensing applications Proc 3rd Int Conf Sensing Technology ICST 2008 Nov 30 Dec 3 2008 pp 192 199 4 V Iyer G Rama Murthy and M B Srinivas Environmental measurement os for a tiny CRF stack used in wireless network Modern Sensing Technol special issue pp 72 86 2008 ISSN 1726 5479 copyright 2006 by IFSA www finebook ir 14 Modeling Sensor Networks Through Design and Simulation Design and programming problems are i
192. ensors can infiltrate people s everyday lives thereby providing real time information about their surroundings Thus online sensor based networks will have tremendous future in many of these applications Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 305 www finebook ir Bibliography Agre J L Clare and S Sastry A taxonomy for distributed real time control systems Adv Computer 49 303 352 1999 Akyildiz I F W Su E Cayirici and Y Sankarasubramaniam A survey of sensor networks IEEE Commun Mag 8 102 114 2002 Akyildiz I F W Su Y Sankarasubramaniam and E Cayirci Wireless sensor networks A survey Comput Networks 38 393 422 2002 P802 11k C LM Amendment to STANDARD FOR Information Technology Telecommunication and Information Exchange between Systems Local and Metropoli tan Area Networks Specific Requirements Part 11 Wireless LAN medium access control MAC and Physical Layer PHY Specifications Radio Resource Measurement of Wireless LANs 1999 Anonymous Optimal data fusion in multiple sensor detection systems JEEE Trans Aerospace Electro Syst AES 22 98 101 1988 Basavaraju S Sensim A Wireless Sensor Network Simulation Template M S Project Dept Computer Science Louisiana State Univ Baton Rouge Bharghavan V
193. ent way to specify constraints of the overall behavior of a system Being in a state where the system responds to only a subset of all allowed inputs produces only a subset of possible responses and changes state directly to only a subset of all possible states Figure 12 2 shows the active objects that constitute a networked sensor embedded system Active objects in a micro framework are encapsulated tasks each embedding a state machine and an event queue that communicate with one another asynchronously by sending and receiving events For active objects within the framework events are processed sequentially in a run to completion RTC fashion while the scheduler encapsulates all the details of thread safe event exchange and queuing REFERENCES 1 C G Cassandras and S Lafortune Introduction to Discrete Event Systems Kluwer Aca demic Jan 1999 2 G S Fishman Principles of Discrete Event Simulation Wiley 1978 3 K Ilgun R A Kemmerer and P A Porras State transition analysis A rule based intrusion detection approach IEEE Trans Software Eng 21 3 151 180 March 1995 www finebook ir 238 INSPIRE 4 V Iyer R M Garimella and M B Srinivas Min loading max reusability fusion classi fiers for sensor data model Proc 2nd Int Conf Sensor Technologies and Applications SENSORCOMM 08 Aug 25 31 2008 pp 480 485 5 V Iyer G Rama Murthy and M B Srinivas Training data compression algorithms and re
194. epartments He is a fellow of 1 the Institute of Electrical and Electronics Engineers IEEE 2 Association for Computing Machinary ACM 3 American Association for the Advancement of Science AAAS and 4 Society of Design and Process Science SDPS and a member of the European Academy of Sciences He has won many best paper awards and IEEE Computer Society awards Dr N Parameshwaran Nandan Parameshwaran is a senior lecturer in the School of Computer Science and Engineering University of New South Wales Sydney Australia He obtained his Ph D from the Indian Institute of Science Bangalore India His research xxi www finebook ir xxii ABOUT THE AUTHORS interests include multiagent system models ontology driven techniques for concept mappings robust rule based systems and network management He has been applying agent based techniques for real world applications including social systems such as road traffic simulations and fireworld scenarios He teaches courses in Internet programming functional programming data structures and algorithms and artificial intelligence Dr Parameshwaran has served as a member in the Florida Artificial Intelligence Research Society FLAIRS and the International Conference on Semantic Computing ICSC programming committees and is a former member of both the IEEE Computer Society and the American Association for Artificial Intelligence AAAI Dr Vir V Phoha Vir V Phoha is a profes
195. er Science Louisiana State University Baton Rouge Mallanda C Sensor Simulator A Simulation Framework for Sensor Networks master s thesis Dept of Computer Science Louisiana State Univ Baton Rouge Michael C and A Ghosh Using finite automata to mine execution data for intrusion detection A preliminary report Lect Notes Comput Sci 1907 2000 66 79 2000 Misra J Distributed discrete event simulation ACM Comput Surveys 18 1 39 65 March 1986 Moitra A and S S Iyengar Parallel algorithms for some computational problems Adv Comput 26 93 153 1987 Nuansri N S Singh and T S Dillon A process state transition analysis and its application to intrusion detection Proc ACSAC1999 1999 pp 378 388 OPNET Technolgies Inc Opnet Modeler www opnet com Park S A Savvides and M B Srivastava Sensorsim A simulation framework for sensor networks Proc 3rd ACM Int DRAFT Workshop on Modeling Analysis and Simulation of Wireless and Mobile Systems 2000 pp 104 111 Perrig A R Szewczyk V Wen D Culler and J D Tygar Spins Security protocols for sensor networks Wireless Networks 8 5 521 534 2002 Polastre J J Hill and D Culler Versatile low power media access for wireless sensor networks Proc 2nd Int Conf Embedded Networked Sensor Systems SenSys 04 ACM New York 2004 pp 95 107 Ramadge P J and W M Wonham Supervisory control of a class of discrete event processes SI
196. er just runs the stack while 1 apsFSM ftendif www finebook ir MOTIVATION AND BACKGROUND 235 12 1 9 Ultralow Duty Cycling Using FARMS When energy resources are abundant due to the availability of renewable energy resources the running harvesting application is sensitive to the recharge rate Since this allows the transmission of data to a forwarding node the communication need not be synchronized The duty cycle still needs to periodically poll the channel to check for any activity For reliable delivery a sufficient number of receivers should be active to ensure that no forwarded packets are dropped This feature needs to be scheduled since receiving and idling can drain all the nodes in a given area if they are not periodically scheduled 12 1 10 Real Time System Components To build a predictable system all of its wireless sensor network components hardware and software plus a good design contribute to this long term predictability Having both good sensor hardware and a good networkable RTOS is a minimal but insufficient requirement for building a correct reactive sensor system An improperly designed testbed system with excellent hardware and software building blocks may still lead to disaster This section deals with sensor testbed microframework In general a good microframework can be defined as one that has a bounded predictable behavior under all system load scenarios 1 The current architecture uses a
197. er of nodes Number of nodes in region 10 A similar scenario was developed in ns2 and the performance of the simulation that is the time taken by the simulator to complete the application was compared in both of them The results show that SensorSimulator takes less time than ns2 even when the numbers of nodes are increased to 2000 The results were validated by confirming that the query nodes are getting back the appropriate data from the region The next simulations are carried out for high traffic scenarios The number of nodes are varied from 500 to 2000 with 100 nodes generating queries at random intervals This result as seen in Fig 14 15 shows that SensorSimulator is able to perform better than ns2 even for high traffic networks Number of queries 100 Simulation time 250 s Network dimension varies with the number of nodes Number of nodes in region 15 The memory used was also compared for both simulators and our observations show that SensorSimulator consumes less space than does ns2 These results presented in Fig 14 16 show that the data structures employed for the simulation are used in a scalable manner to represent the different classes and the interaction with the OMNeT framework It can also be observed that the rate of memory usage www finebook ir 12000 10000 8000 6000 Time sec 4000 2000 0 FIGURE 14 15 FINAL COMMENTS 271 SensorSimulator 2 ns2
198. ere will be keys in common between nodes in different subgrids After being randomly deployed the nodes discover common keys authenticate and communicate securely The analysis and simulation results show that this scheme is able to achieve better security compared to the random schemes For a full treatment on this topic refer to the article by Kalindi et al 3 16 5 SECURE INFORMATION ROUTING Nodes deployed in hostile environments are prone to capture Capture of a single node discloses all the information about the keys that they contain More specifically www finebook ir PROBLEMS 303 an adversary can capture multiple nodes by eavesdropping on radio transmissions injecting bits into the channel and repeating previously heard packets Adversary nodes can be about as powerful as existing nodes or significantly more powerful The paper by Karlof and Wagner 4 presents a threat model based on two classes of attacker mote class attacker where the attacker has access to a few sensor nodes similar to legitimate nodes and laptop class attacker where adversaries have access to more powerful computational power more battery power and a high powered radio transmitter This algorithm presents a secure routing protcol that guarantees protection against eavesdropping integrity authenticity and availability of messages For more details refer to Ref 4 16 6 SECURITY PROTOCOLS FOR SENSOR NETWORKS This particular algorithm has man
199. ere we sense the world using the sensor mounted on the mote and turn on the LED if the data we read have a 1 as the www finebook ir 120 SENSOR PROGRAMMING least significant digit This contains a module SenseC an interface Read and a configuration SenseAppC Courtesy of TinyOS authors Jan Hauer description Sense Application label prog7 module SenseC i uses interface Boot interface Leds interface Timer lt TMilli gt interface Read lt uint16_t gt implementation event void Boot booted call Timer startPeriodic 10 event void Timer fired call Read read event void Read readDone errort result uinti6 t data 1 if result SUCCESS if data amp 0x0001 call Leds led0OOn else call Leds ledOOff SenseC This module uses four interfaces three of which we have already seen Boot Leds and Timer See code above The fourth interface Read is described below Read This interface provides one command function and one event function to read the sensor and return the value read interface Read lt valt gt command error_t read event void readDone error_t result val_t val www finebook ir SENSING THE WORLD 121 The command read is invoked for the read operation We can access the results of the read command by calling the event function readDone result val When we call as readDone result val we can retrieve the value of the data
200. ernet 21 5 1027 1031 1991 V Iyer S S Iyengar N Balakrishnan V Phoha and M B Srinivas Farms Fusionable ambient renewable MACs Proc Sensor Application Symp Feb 2009 A Moitra and S S Iyengar Parallel algorithms for some computational problems Adv Comput 26 93 153 1987 J Polastre J Hill D Culler Versatile low power media acess for wireless sensor networks Proc 2nd Int Conf Embedded Networked Sensor Systems SenSys 04 ACM New York 2004 pp 95 107 I Rhee A Warrier M Aia J Min and M L Sichitiu Z mac A Hybrid MAC for Wireless Sensor Networks IEEE Press Piscataway NJ 2008 vol 16 pp 511 524 C Xavier and S S Iyengar Introduction to Parallel Algorithms Wiley 1998 W Ye F Silva and J Heidemann Ultra low duty cycle MAC with scheduled channel polling Proc 4th Int Conf Embedded Networked Sensor Systems SenSys 06 ACM New York 2006 pp 321 334 www finebook ir 9 Techniques for Protocol Programming The function of good software is to make the complex appear to be simple Grady Booch Wireless sensor networks are made up primarily of spatially distributed autonomous devices that use sensors to cooperatively monitor certain physical entitle such as temperature sound vibration and pressure at different locations Most sensor nodes are used in one of two contexts as distributed databases or for event detection 2 Irrespective of the r
201. es as well as the number of edges Consider the planar graph G drawn in a plane without its edges intersecting The graph divides the plane into several regions Figure 4 44 shows a planar graph and the region R R2 Re It is easy to observe the following results 1 Any simple planar graph can be embedded in a plane such that every edge is drawn as a straight line segment a A planar graph can be embedded in a plane such that any specified region can be made the infinite region Ks K3 3 FIGURE 4 43 Kuratowski s graphs www finebook ir COLORING AND INDEPENDENCE 83 FIGURE 4 44 A planar graph with six regions b Any planar graph can be embedded in a sphere and any graph that can be embedded in a sphere is a planar graph 2 If G is a connected planar graph with n vertices and m edges it then would generate m n 2 regions For example in Fig 4 44 n 13 m 17 and the number of regions is m n 2 17 13 2 6 Also we can prove that for a planar graph m 3n 6 This is a necessary but not sufficient condition Kuratowski s second graph is a nonplanar graph satisfying this condition A necessary and sufficient condition for a graph G to be planar is that G does not contain either of Kuratowski s two graphs or any other graph homeomorphic to either of them An outer planar graph is a planar graph that has a planar embedding in which all the vertices of the graph lie on the exterior regio
202. es be tween a few kilohertz and several megahertz allowing for the most basic tasks such as sensing and intermediate routing of data from other sensors 8 1 2 Transceiver Currently information can be transmitted in wireless sensor networks through three types of media Optical communication laser Infrared IR communication Radio frequency RF Although both optical and IR communications have the advantage of requiring less energy to transmit data they both suffer from major drawbacks For instance optics based communications require both communicating nodes to be properly aligned in direct line of sight Conversely IR communications have very short ranges Other issues include sensitivity to atmospheric conditions in optical communications and unidirectional communication for both systems It is for these reasons that RF is more prevalent in sensor networks 8 1 3 Memory As noted in earlier chapters memory is a scarce resource on sensor nodes Therefore applications must be efficient in terms of not only energy but also the amount of memory they use For instance a mica2 mote contains 644 kb of memory that must be used by the TinyOS operating system and applications built on the platform 1 8 1 4 Power Unit Most power consumption in sensor nodes is primarily for data processing sensing or communication operations Of these operations communication activities exert the most toll on the battery life of the sensor
203. es to another component thereby allowing two or more components to communicate with each other 3 5 2 AN INTRODUCTION TO NesC In TinyOS the software component abstractions modules and configurations 3 are written in a dialect of the C language called nesC NesC contains a subset of features of the standard C libraries and syntax with some extensions such as commands and events added to accommodate its event driven approach to programming rather than the traditional imperative style of many C based languages Due to the resource constrained environments typical of most sensors these additional features in nesC help improve the efficiency of programs written in the language These new constructs are outlined in Figs 5 1 5 3 www finebook ir 94 TINY OPERATING SYSTEM TinyOS TinyOS FRAMEVVO FIGURE 5 1 TinyOS system services Tire Interface Engine in Interface pos Interface CN F w p IG FIGURE 5 2 Three sample modules and their associated interfaces FIGURE 5 3 Acar configuration consisting of multiple modules www finebook ir AN INTRODUCTION TO NesC 95 52 1 Split Phase Operations Commands and Events In addition to the standard C function declaration functions can be declared as commands events or tasks command return type gt lt command name gt lt input parameters gt command error_t readSensor event void lt event name gt lt returned messages and errors gt
204. es well with the number of nodes We have tested between 100 and 500 nodes set up randomly in a 420 x 420 m grid Its strength lies in this scalability as well as its exceptional radio prop agation model GlomoSIM was chosen as the base simulator because its framework www finebook ir SIMULATION CAPABILITIES 249 ENNH Rx END 1200 1000 co c c T a c ce T c e T Lifetime in simulation seconds before the sensor runs out of battery 200 El 0 mJ J B MAC NO SLEEP B MAC SLEEP Qos FIGURE 13 13 Scheduling idle with sleep and low power listening increases lifetime by 2 times is superior to that of NS 2 and other network simulators GlomoSIM allows new models to be added to any layer with ease as shown in Fig 13 9 The basic means of communication for multihop or internode communication is their message passing APIs An application generates a message to be sent out of the radio using the same API as it uses to set a timer for itself This independence of the message passing API from the simulation leads to simpler programming The strength of Glomo SIM is its scalability and exceptional propagation models This is important because of two factors 1 Scalability When simulating sensor networks it is not uncommon to simulate tens of thousands of nodes While a traditional network might be simulated with only 100s of nodes a sensor network is inherently larger in size A sim
205. eveloped and is not available www finebook ir 256 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION OPNET Modeler is a commercial platform for simulating communication net works 17 Conceptually the OPNET model comprises processes that are based on finite state machines and these processes communicate as specified in the top level model The wireless model is based on a pipelined architecture to determine con nectivity and propagation among nodes Users can specify frequency bandwidth and power among other characteristics including antenna gain patterns and terrain models J Sim is another object oriented component based discrete event network sim ulation framework written in Java 21 Modules can be added and deleted in a plug and play manner and J Sims is useful for both network simulation and emulation by incorporating one or more real sensor devices This framework provides support for target sensor and sink nodes sensor channels and wireless communication channels physical media such as seismic channels power models and energy models GloMoSim is a collection of library modules each of which simulated a specific wireless communication protocol in the protocol stack 25 It is used to simulate ad hoc and mobile wireless networks 14 3 4 The OMNeT Framework Objective Mocular Network Testbed in C OMNeT is a public domain component based modular simulation framework 23 It is has been used to simu
206. event void readDone error_t result uinti6 t value Commands are functions already implemented by the called software component in which its invocation and completion paths have been disassociated In other words most commands when called return immediately without completion and if any errors or messages result at the end of its computation a callback function is signaled The notion of having a separate callback function signaled after the completion of a command is called split phase This is particularly useful since TinyOS contains no blocking operations therefore any long running operation can immediately return control to the calling program and on completion a callback signaling the event associated with the command function is signaled A command and Its Associated Callback Function Events are triggered in response to some action this may be after the execution of a command or in response to messages from the environment For example the reception of a packet triggers an event receive Typically events are not implemented by the called component but rather are functions whose implementation details are left to the calling software component Simply put events are methods that must be overridden by the caller All callback functions signaled after the termination of a command are events in which case their role would be to process the result of the terminated command Code Sample Illustrating the Split Phase Concept In the block
207. ewed Binary tree Right skewed Binary tree FIGURE 4 24 Skewed binary trees node are usually called the eft child and the right child A binary tree in which no vertex has a left child is said to be right skewed We define a left skewed binary tree similarly Figure 4 24 shows left skewed and right skewed binary tree Consider the binary tree shown in Fig 4 25 At level 0 the root alone is there At level 1 there are two vertices At level 2 there are four vertices and at level 3 there are eight vertices This observation leads to the following theorem Theorem 4 1 1 In a binary tree there are at the most two vertices at level i 2 The maximum number of vertices in a tree of height k is 2 Proof The proof of condition 1 in Theorem 4 1 is on induction on the level number and the result is true for i O If the result is true for a level i there are at most 2 vertices at level i Each of those vertices can produce at the most two children for level i 1 So the maximum number of vertices at level i 1 is 2 1 Hence the result of condition 1 follows by induction on i The proof of condition 2 is based 12 13 14 15 FIGURE 4 25 A full binary tree www finebook ir 70 DATA STRUCTURES FOR SENSOR COMPUTING on the following numerical result 29 d 0 a Pa pepe a T 4 2 A tree of height k with all the 2 1 nodes is called a full binary tree In a full binary tree nodes can be numbered serially as follows The roo
208. ex is left unvisited There are two methods for traversing the vertices 1 Breadth First Search BFS and 2 Depth First Search DFS Let v be the starting vertex for traversing the graph G The BFS starts by visiting the vertex v After visiting v all its adjacent vertices are visited in some defined order Let the adjacent vertices of v be v v2 vg After visiting these vertices we must www finebook ir 76 DATA STRUCTURES FOR SENSOR COMPUTING FIGURE 4 34 A graph visit all the unvisited vertices that are adjacent to these vertices The procedure is repeated until all the vertices of the graph are visited Inthe DFS technique we start from v Then we visit an adjacent vertex v of v Now we visit an unvisited vertex adjacent to vi At some stage if u has no unvisited vertex adjacent to it we backtrack and come to V _ and visit an unvisited vertex adjacent to it The procedure is repeated until we backtrack to v and visit all its adjacent vertices The traversal forms a spanning tree for G A graph G its BFS spanning tree and its DFS spanning trees are shown in Figs 4 34 and 4 35 In the BFS spanning tree the dashed lines represent the chords and the solid lines the branches In the BFS spanning tree note that every chord joins vertices either at the same level or at consecutive levels This property can be effectively used to solve several problems in graph theory 4 10 CONNECTIVITY If G V E isa graph and v V G
209. f split phase operation We will present concepts in the nesC language using a sequence of simple examples where we will attempt to explain one chosen concept at a time 6 2 A SIMPLE PROGRAM A typical nesC program henceforth referred to as a component consists of an inter face a module or several modules and a configuration file showing the connections between the various components of the modules An interface consists of a set of clearly defined commands and event signatures that represent how interactions can Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright 2011 John Wiley amp Sons Inc 99 www finebook ir 100 PROGRAMMING IN NesC occur between multiple modules 1 An example showing how a simple PrintMes sage interface is implemented is shown below PrintMessage nc This interface called PrintMessage consists of a command send and an associated event send Done Any component making use of this interface can call the command send with the appropriate parameters and expect the corresponding sendDone event to be signaled if there are no errors during execution The actual implementation of the command send is specified in the PrintMessageC module shown below By design commands are implemented in the called components while the caller handles the implementation of the events interface PrintMessage
210. f you have seen existing design and codes that are designed for various parts of the network stack you may have observed that they are riddled with a disproportion ate number of convoluted conditional execution branches deeply nested if else or switch case statement in C C This highly conditional code is a testament to the basic characteristics of reactive systems If a redesign could eliminate a fraction of these conditional branches the code would be much easier to understand and test and the sheer number of convoluted execution paths through the code would drop radically perhaps by orders of magnitude Techniques based on state machines are capable of achieving exactly this dramatic reduction of the different paths through the code and simplification of the condition tested at each branching point 3 7 The state machines described in the cluster timing UML specification represent the current state of the art in the long evolution of these techniques 12 2 2 Sensor State Machine UML Diagram Algorithm A system exhibits state behavior when it operates differently during different periods and its behavior can be partitioned into finite nonoverlapping chunks called states For example basic mathematical functions such as sin x return the same result for a given input x regardless of the history of previous inputs x A common straightforward way of modeling state behavior is through a Finite State Machine FSM Using FSMs is an effici
211. fer to Using the Global Sensor Network GSN reference provided install and con figure the GSN middleware What is the primary difference between server tier tools and client tier tools List three other examples of data aggregating platforms such as GSN REFERENCES N 00 10 Q PWN 10 Crossbow Imote2 Datasheet courtesy Crossbow Inc Crossbow MIB520 Datasheet Courtesy Crossbow Inc Crossbow Moteworks Software Reference Manual courtesy Crossbow Inc Crossbow Product Feature Reference Manual courtesy Crossbow Inc Crossbow TelosB Datasheet courtesy Crossbow Inc Global Sensor Networks GSNTeam http blog xbow com xblog sensorboards http inst eecs berkeley edu cs194 5 sp08 lab 1 index html Research Integration Platform Survey Embedded WiSeNts consortium M Ruiz Sandoval T Nagayama and B F Spencer Sensor development using Berkeley mote platform J Earthquake Eng 10 289 309 2006 www finebook ir PART II Background www finebook ir 4 Data Structures for Sensor Computing Understanding the fundementals of data structures is essential to writing efficient code S S Iyengar Sensor computing consists simply in manipulation of sensor data in a suitably chosen data structure for event data driven sensor computing applications 5 It can also refer to the design and analysis of sensor data abstractions An in depth understanding of the structural properties of certai
212. file getdata line 2 file amp amp strcmp getdata x 3 file 04 results results 1 end end sensor visualization 10 getdata line 2 file 0 NOVO event type if mod results 2 disp Bag 4 getdata line 2 file went into temp storage l into storage time checked in time random default have not handled oe case where bag was never checked in for x 1 line if strcmp getdata x 2 file getdata line 2 file amp amp strcmp getdata x 3 file 01 tmp datevec getdata x 1 file time checked in datenum tmp2 tmp 4 6 1 end end times from checkin to storage times from checkin to storage time time checked in else disp Bag 4 getdata line 2 file went out of temp storage out of storage disp Bag getdata line 2 file W loaded_on van getdata line 4 file sensor visualization 13 getdata line 2 file 0 getdata line 4 file event type 2 case 06 check if bag that failed security scan was put on plane results 0 for x 1 line www finebook ir 290 MATLAB SIMULATION OF AIRPORT BAGGAGE HANDLING SYSTEM if strcmp getdata x 2 file getdata line 2 file amp amp strcmp getdata x 3 file 03 amp amp strcmp getdata x 4 file 0 results results 1 end end if results gt 0 disp Bag getdata line 2 file loaded onto flight 4 getdata line 4 file sensor visualization 11
213. fined as how the stack performs load balancing reusability index power aware sleep scheduling due to network density and the reliability of sending sensed data wirelessly at the datalink layer 2 Reusability Index This performance based index can described as the number of times that a given node has been used as a clusterhead to communicate to a base station or a sink during its lifetime As many of the clusterhead selection algorithms are distributed in nature they will not overuse a specific node more than the critical number of times If all nodes are used evenly then the reliability of the network increases during the entire lifetime of the node 3 Sleep Scheduling Most of the deployed sensor network applications are dense because of the limited radio transmission range so even when not transmitting data nodes are subjected to overhearing and collision These factors severely impact the total power consumed So in a dense deployment if a sufficient number of nodes are awake to receive the multihop traffic then other nodes can shut off their radios after exchanging the next polling time to minimize idling By activating only a subset of nodes and scheduling timeslots for nodes to be active sleep scheduling saves on power and avoids dropped packets The end goal of each of these methods is to continue reciving data from the network for as long as possible www finebook ir PERFORMANCE DRIVEN NETWORK SOFTWARE PROGRAMMING 9 4
214. fusionable ambient renewable access control applications We would like to thank the graduate students involved in proofreading editing including Robert Rajesh Srivathsan Srinivasagopalan and many others The authors would like to hear from any readers with comments suggestions or bug reports In writing a book of this proportion it goes without saying that many people have contributed both directly and indirectly to the inception development and completion of this project This endeavour would not have been achieved if not for the many hours toiled by both my present and former students For this thanks goes to Professors N S V Rao Oak Ridge National Lab Richard Brooks Clemson Mengxia Zhu SIU Carbondale and Qishi Wu U Memphis and current PhD students Robert Dibiano Srivathsan Srinivasagopalan Vasanth Iyer J Kim Further I want to thank the faculty at KAIST S Korea specifically Dr Song and his collaborators who participated in the early discussions of this project I want to thank all of my research collaborators specifically Krishnendu Chakrabarty Duke Sartaj Sahni Florida Bhaskar Krishnamachari Southern Cal xix www finebook ir XX ACKNOWLEDGMENTS and at Louisiana State University Professors S Mukhopadhyay R Kannan C Busch J Zhang H Wu X Li Moreover We Iyengar and Phoha would like to thank various funding agencies specifically DARPA ONR NSF Army Research Office Air Force A
215. g Cambridge Univ Press 2009 www finebook ir 6 Programming in NesC Low level programming is good for the programmer s soul John Carmack 6 1 NesC PROGRAMMING Because of the highly resource constrained nature and unpredictability associated with wireless sensor networks traditional programming techniques cannot be used in the development of programs intended for use in these networks Unlike conventional programming techniques in which most resources are available locally and remote resources could be accessed through persistent connections sensor applications have to contend with several factors that may lead to degradation in performance of the sensor network ranging from energy depletion to limited or no connectivity to physical damage of the sensor hardware in hostile environments In TinyOS based sensor networks a C based language called nesC has been developed with several constructs built in to mitigate some of these undesirable properties of sensor networks NesC shares a similar syntax with C but runs only on the TinyOS operating system The language is simple so that it can be used for programming embedded systems wireless sensor nodes and similar applications where efficiency in terms of speed and memory is of utmost importance The nesC language has features borrowed from computer hardware such as the notion of components and subcomponents each interacting with the other via clearly defined interfaces and the concept o
216. g h E a b b c c d d a te f 9 g h h e a e b c g d h It is preferable to draw a graph on the two dimensional 2D plane so that no two edges intersect each other It is evident that we cannot draw all the graphs in the plane in such a way that no two edges intersect A graph is said to be planar if it can be drawn in a 2D plane without its edges intersecting Figure 4 42 shows some simple examples of planar graphs A graph G V E is said to be bipartite if V can be partitioned into V and V5 such that every edge of G joins a vertex of V and a FIGURE 4 41 Same graph different schematic representation www finebook ir 82 DATA STRUCTURES FOR SENSOR COMPUTING FIGURE 4 42 Some planar graphs vertex of V2 Note that a bipartite graph contains no cycle of odd length A complete bipartite graph is a bipartite graph in which every vertex of V is adjacent to all the vertices of V2 A complete bipartite graph with m vertices in V and n vertices in Vz is denoted by K Note that K has m n vertices and mn edges The Polish mathematician Kazimierz Kuwatowski 1896 1980 proved that the following two graphs see also Fig 4 43 are nonplanar 1 The complete graph with five vertices is given by Ks 2 The complete bipartite graph K3 3 Observe that both Kuratowski graphs are regular and nonplanar Also note that they are minimal nonplanar graphs with respect to the number of vertic
217. g e speq Jo 1equinN e sjeDuessed jo 1equinwN e ponad xeag e ulejs s jo indyBnosu pue yoedeo uo sjooojoJd uonnjosaei wieje jo pedu puejsjepufi uonnjios jeuundo pue wajsAs uo SO EI BOIAIOS juaJojllp jo 19eduui pue si pun ulejs s eui jo Ayoedeo senenb Ul 49euo uo sejeJl o IAJ8s Jejunoo 49491 pue sjejeurejed pue pueuiep noy yeed uBisep Aey euiuuejeq jo pedu puejsjepuf Sje e puewep eBeb6eq pue JeBuessed Ajrep pue Jnoy yeed eui euiuuejeq uoneuos uonnjosed uue v Buiueegog Ajunoes Buinenp ul yo8yD pueuieq eBeb6eg pue 1eBuesseg SOLayy Bong sjeo suoijoun4j 280 www finebook ir jeroure pue ppuru jj0dure uooAjoq uodsuen asesseg PST WH 5LI uooous moj eBeBBeq dd U9UJ8AOUIJ JAYA lt uiuo oz uiujoo OL WA uiuo OL XX U UbI GL A e uu KY suey a E 5 T T 2 wa ua m um um uuujoo oz o W yya ul y9949 3uniuvdaa w s umos TVAIHHV 281 www finebook ir ssooo1d Sur pueq eSe33eq oy Sunipodxo 10 onpa posodoig I AMANDA uooous Mo eBe6Beq lt n huaweAow e otuaA 3 UILU OD OZ U U 08 A u u 08 Af lt ie jee eas oz V EMO ERA uluu o5 9 PA 9 a20 ywy 0g ul y0949 3uniuvdaa uiejs s Bunios TIVAIHHV 282 www finebook ir SIMULATION RESULTS AND DISCUSSION 283 zu Los Dus zu Lor Dur
218. gins its execution at the event function Booted booted As we trace down in the figure the reader may also want to verify the control flow in the program module Boot booted invokes PSControl start to turn on the transmission receiver hard ware component which attempts to turn on the component and returns the results in the parameters of the event function PSControl startDone The event function startDone checks to see if there has been any error in starting the device If there was any error we go back and call the start function again However if the starting of the device was successful that is error success we start the clock by calling www finebook ir APPLICATIONS USING THE INTERFACE SplitControl 127 Timer0 startPeriodic which invokes the event function Timer0 fired See imple mentation for event PSControl startDone below As soon as this happens in the function fired we turn the LEDs by setting the bits to 0011 on showing that we were able to successfully start the transmission device See left branch in Fig 7 5 Then we proceed to send a packet calling sendPacket by first constructing the packet calling Packet getpayload and then sending it by calling AMSend send The command send upon completion of its task invokes the event function AMSend sendDone which verifies if the packet was successfully sent See code for event void AMSend sendDone below We then go back and turn on the LEDs to
219. gm in which program flow is determined by events 2 3 In wireless sensor networks where conservation of computational and energy resources is paramount event based programming is very critical In the absence of an event driven approach programs running on these sensing devices would have to continuously poll the system checking for the occurence of certain events This approach would be both wasteful of system resources and highly ineffective in larger sensor networks causing program degradation In TinyOS the event driven programming approach allows for fine grained control over power usage while maintaining the scheduling flexibility required in sensor networks due to their unpredictable nature The following code snippet shows how events can be triggered and handled in a nesC program if call AMSend send packet sizeof Data event message t Receive receive message t msg void pay load uint8_t len 1 call Leds ledOToggle busy FALSE busy TRUE www finebook ir REFERENCES 97 In this code a node sends a data packet using its radio and the send interface while setting the busy flag to true The data sent by node 1 triggers a corresponding receive operation on all nodes within the transmission radius of the first node without the need for continuous polling For a more complete description of TinyOS programming the text by Levis and Gay see 3 PROBLEMS 5 1 Briefly explain the components of
220. gned dif ferent addresses 9 7 Describe content based addressing Explain with examples how and where this addressing scheme would be useful 9 8 What is flooding Describe its use and drawbacks in the context of sensor networks 9 9 What are some of the improvements over fooding Explain any two REFERENCES 1 V Bharghavan A Demers S Shenker and L Zhang Macaw A media access protocol for wireless LANs Proc ACM SIGCOMM 1994 1994 2 C G Cassandras and S Lafortune Introduction to Discrete Event Systems Kluwer Aca demic Jan 1999 3 J Hill R Szewczyk A Woo S Hollar D Culler and K Pister System architecture directions for networked sensors in Jn Architectural Support for Programming Languages and Operating Systems 2000 pp 93 104 4 W Ye F Silva and J Heidemann Ultra low duty cycle MAC with scheduled channel polling SenSys 06 Proc 4th Int Conf Embedded Networked Sensor Systems ACM New York 2006 pp 321 334 www finebook ir PART IV Real World Scenarios www finebook ir 10 Sensor Deployment Abstraction Simplicity and elegance are unpopular because they require hard work and discipline to achieve and education to be appreciated Edsger Dijkstra As we move into the future the effects of Moore s law will progessively make the unit cost of sensor devices more and more negligible Indoor and even outdoor environments with access to the existing power grid will be fertile e
221. gnoring it will lead to incorrect trends in power consumption in the network New processor models with enhanced features and improved energy consumption levels can be incorporated in this module for testing various kinds of applications The CPU Base abstract class forms the basis for different CPU models and defines the interfaces of this module with the coordinator and the battery CPU Simple implements the power consumption of the CPU in different states idle sleep and active 3 Radio Model This model is used to characterize the antenna property of a node RadioBase is an abstract class for the different radio models Radio Simple a subclass of RadioBase updates the energy of the battery depending on the state of the radio idle sleep transmit and receive The values for the different properties of the hardware and consumers may be provided through the configuration file 14 4 3 Wireless Channel Model The wireless channel module controls and maintains all potential connections between the sensor nodes These static connections are provided from all the nodes to the wireless channel module and from the module to all the nodes in the NED file These connections enable sensor nodes to exchange data and communicate with each other Any message from a node is sent to all the neighbors within its trans mission region with a delay d where d is distance between communicating sensor nodes speed of light Various radio propagation models
222. h as 10 samples per second sps 13 4 3 Channel Emulation The MAC and radio layer use the modified version of the GlomoSIM radio This is very useful as it combines the best of simulations with the best of the real world test bed 13 5 BENCHMARKING Table 13 3 and Figures 13 13 and 13 14 present a detailed result of programmable simulations of cross layer structures of sensor networks more specifically the results highlight the time scheduled by MAC during useful network lifetime For more details on implementation refer to Ref 1 For futher details on benchmarking refer to Refs 1 and 2 13 6 CONCLUSION The fundamental contribution of this chapter is in providing a reprogrammable structure in developing energy aware routing for scalable sensor networks see also www finebook ir 252 PERFORMANCE ANALYSIS OF POWER AWARE ALGORITHMS Figs 13 13 and 13 14 Computational networking characteristics of INSPIRE in clude effective link abstraction sensor service allowing network protocols to run efficiently on varying power management schemes power savings greater simpler code multiple network protocols that benefit from coexistence coordination and co operation effective separation of mechanism and policy building blocks for a sensor network architecture and potential application to internet architecture and 802 11 PROBLEMS 13 1 List and explain in your own words four factors that degrade the performance of a wireless s
223. h example is Crossbow s KAKI FBoH a L rien 7 FIGURE 3 11 The SHIMMER platform www finebook ir TABLE 3 5 Gateway Device Properties SERVER LEVEL 35 Feature Stargate net bridge MIB520 MIB600 Program flash 8 MB RAM 32 MB IO Extra features 1 x RJ45 and 2 x USB 2 0 USB interface RJ45 USB flash disk 2 GB JTag interface for POE ARP DHCP onboard server debugging Telnet Size 130 x 21 x 91 mm 4 63 x 2 29 x lin FIGURE 3 13 The MIB600 programming board www finebook ir 36 SENSOR TECHNOLOGY XServe 3 software from their Moteworks development suite of applications based on TinyOS XServe acts as a primary server running on a PC or gateway device from which the data gathered from the sensor network are interpreted Other examples include the Global Sensor Network GSN middleware 6 3 3 CLIENT LEVEL The client level consists of all publishing visualization and monitoring applications A number of powerful platforms exist today that support advanced publishing of sensor data Global Sensor Network middleware supports publishing sensor streams to RSS feeds security management system SMS text updates Web publishing and several other applications Other examples include the Mote View from Crossbow 3 34 PROGRAMMING TOOLS Setting up and configuring your sensor programming environment can be a daunting task if there is in adequate knowledge of a
224. he bridge The design of such nodes needs to find a new way to integrate MEMS complementary metal oxide semiconductor CMOS and radiofrequency RF technologies Researchers at University of California Berkeley have developed a prototype of system on chip node called Spec Mote The mote combines ultra low power computation communication and sensing into a small only 5 mm 2 x 2 5 mm single chip Networked sensors broaden our capability to observe the physical world Sensor networks enable us to observe objects at close range and provide the possibility of monitoring previously unobservable phenomena Satellites and radars are large and powerful sensors that have been widely used This type of sensor has long range sensing ability and a single sensor can detect or monitor a large area However such sensors are not suitable for an environment where the line of sight paths are very short for example detecting individual animals in a forest or monitoring a patient in a hospital Small sensors can be used in such situations Since each sensor node has a short sensing transmitting range and limited processing ability a number of nodes are networked to accomplish complicated tasks As the sensor nodes can be placed close to or in the objects greater informational accuracy is achieved and some previously unknown phenomena may be discovered Additionally large numbers www finebook ir SENSOR NETWORK APPLICATIONS 17 of networked sensors can i
225. he same simulation results obtained on both the simulators During the simulation runs we measured the memory allocated before the start of the simulation that is giving the memory usage for the initialization and the setup of the objects of the simulation The memory usage during the simulation was also measured The results for the memory usage are as shown in Fig 14 10 and in Fig 14 11 for 10 nodes sending queries Figs 14 12 and 14 13 show the performance of the simulators for 100 nodes sending queries to the region This shows that the data structures used for the simulation are used in a scalable manner to represent the different classes and the interaction with the framework It can also be observed that the rate of memory usage increases at a rate that is faster for ns2 than for SensorSimulator thus allowing for large simulation setup and more scalability in SensorSimulator than in ns2 600000 500000 E 2 400000 gt SensorSimulator 300000 E ns2 9 200000 100000 0 7 e N N N Number of Nodes FIGURE 14 10 Memory consumption before simulation starts 10 queries www finebook ir EXPERIMENTAL RESULTS 269 600000 _ 500000 E 2 400000 radi SensorSimulator 300000 e Fi c ns2 E 200000 z p 100000 ee 0 Seo T S S S S S S SS S SS SS Number of Nodes FIGURE 14 11 Memory con
226. he specifications and functionality of the MTS400 multisensor board Write a one page summary of the article by Akyildiz et al 4 Other than those discussed in this introductory chapter list three advantages and three disadvantages of sensor networks Crossbow Technology Inc s MICAz mote and Europe s Smart Its platform are two popular sensor platforms Research and contrast the features of MICAz with smart Its in terms of size weight battery life onboard sensors memory CPU operating system processing limits radio range etc www finebook ir 1 7 1 8 1 9 1 10 1 11 1 12 1 13 1 14 1 15 1 16 PROBLEMS 13 What are the unique characteristics of programming environments for sensor network software Other than sleep scheduling give two techniques to conserve the battery life energy of nodes in a sensor network State the issues to be considered when updating computation communication software in a sensor network Does the data aggregation strategy adopted by a sensor network application affect its operational integrity and security If Yes explain how and if No explain why In about five paragraphs discuss any three of your favorite real world sensor network applications Sensors mounted on moving objects can open many interesting real world applications For example sensors are already mounted on devices such as mobile phones to sense temperature motion and other parameters
227. hed WSN data aggregation allows reduction of the redundancy in a transmission by statistically evaluating the frequency of occurring samples and the trend direction that they have during its lifetime When sensor nodes sample individually only the aggregated data are transmitted thus increasing the local processing and decreasing the radio usage per aggregation cycle A simple example is using data compression at the nodes to send fewer bits during each transmission 1 4 3 Security Security is a constant threat to outdoor wireless environments thus it is prudent to have an encryption algorithm that allows encryption and decryption of wireless communications One novel way to implement a security algorithm is to have a oneway function which is NP complete at the predeployment stage and cannot be decrypted with limited resources in a deployed site of operation This method is more suitable in other applications of networks in the case of WSN networks because of the nature of their distribution one can design a network polynomial key that is not a local function The broadcast message cannot be decrypted when a few nodes are compromised as it needs to have other parameters that are well distributed and concealed from the intruder www finebook ir 10 INTRODUCTION 15 UNIQUE CHARACTERISTICS OF PROGRAMMING ENVIRONMENTS FOR SENSOR NETWORKS Sensor networks differ from both wired and wireless computer networks in many ways The topology of sens
228. her information process it and interact with each other in networks and how these networks interact with the physical world One aim of this book is to provide fundamental information necessary to write efficient sensor network software A second aim is to provide a balance between theory and applications so that the subject matter is complete self contained Wireless sensor network WSN applications may consist of diverse sensors with varying capabilities For example sensors may range from an extremely constrained 8 bit mote to less resource constrained 32 bit microservers These sensors may be organized in different network configurations which use different communication and data dissemination protocols most software development platforms consist of libraries that implement message passing interprocess communication IPC primi tives tools to support simulation emulation and visualization of networked systems and services that support networking sensing and time synchronization Given all of this diversity there is an underlying theme of software development and deployment that cuts across platforms 11 SOME FOUNDATIONAL INFORMATION This section provides some basic information necessary for understanding the sensors and sensor networks 1 1 4 Sensors Typically a sensor is composed of components that sense the environment process the data and communicate with other sensors computers A sensor responds to a physic
229. hile have address or k lt kmax Get address for self t randomly select an address from T f randomly select an address from F www finebook ir 178 TECHNIQUES FOR PROTOCOL PROGRAMMING Announce If any one has f as his address message me at flood t address t wait ack delta t for delta t units of time if ack Yes I have k and continue else my address f have address YES I have an address now My job from now on is to reply to other s queries for i 1 i lt m i forward m times How to choose m receive requests from other nodes packet p t extract temporary address from p f extract fixed address from p if f my address send to address t my address PERKINS SOLUTION Assume n total number of nodes S0 generate the pair lt t f gt send to all neighbours goto s1 S1 receive receive all acks if all are ve assert f as own address if any is ve goto s0 if it is a request for verifying some other node s request then extract f t gt par t and check if f it clashes with its own address if it does send ve ack to t if it doesn t send ve ack to t nesC code total number of nodes www finebook ir CONTENT BASED ADDRESSING 179 T array of address global n T booted call generate generate generate t f gt send to all neighbours rec
230. hoodtable if flag 1 display Already forwarded else receive jfrom jneighbour query packet p sender P extract 4idof sender p I already have a pointer to B neighbour B lookup jlocal jmemory send packet p to node B wait for reply from node B reply packet q send received reply to P reply packet q Node D Destination node D that reports event e receive jfrom jneighbour query packet p sender P extract jid jof sender p extract original 4 sender node S from packet p new packet p sender S event e send packet p to P Each abstraction specifies an elementary behavior that we find useful in our sensor programming experience All these abstractions will be implemented using the S MAC schedule based listening and sleeping Thus far we have been looking at examples where nodes have to transmit and receive data in a general setting Nodes do not know about each other However in many real world situations nodes have partial knowledge about the each other The following code illustrates one such technique where nodes exchange information more efficiently under certain assumptions www finebook ir QUERYING IN RUMOR ROUTING 193 The program above shows the code for each node involved the querying operations for rumor routing For compactness the roles performed by each of these nodes as specified in the code can be merged together and installed on each
231. ible to embed it in larger applications OMNeT 4 models are built with NEtwork Description NED and omnetpp ini and do not use scripts which makes it easier for various simulations to be configured e e The following sections give the detailed implementation of our simulation scenario on OMNeT 14 4 SIMULATION DESIGN This section presents the architecture of a sensor node and the overall design of our new simulator 6 10 15 The topology of the sensor network field in our simula tions is derived from the simple and compound module concept of the OMNeT4 framework As shown in Fig 14 1 layers of a node behave as simple modules and a sensor node behaves as a compound module and all these sensor nodes constitute the sensor network depicted as a system module The architecture of a sensor node is depicted in Fig 14 2 Each layer of the sensor node is represented a simple module of OMNeT4 The layers communicate with each other through gates and each layers has a reference to the coordinator The structure of a layer is represented in Fig 14 3 These simple modules are connected according to the layered architecture of a sensor www finebook ir 258 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION Sensing Application Network Layer MAC Layer Physical Layer oud PZ C Umoon Sensor Node Direct communication Sensor Channel Messages Wireless Channel Target Object FIG
232. icle collision Deploying sensors in the pavement can assist with traffic control of pedestrians and vehicles 2 1 5 Smart Environments Deploying sensor networks in indoor environments such as buildings homes offices and laboratories can make the environment seem alive Such smart environments can track and record the activity of people in that area or actively react and interact with people Examples of research conducted in this area include home automation residential laboratory building environmental control interactive museum and smart kindergarten applications Embedding smartsensors in structures such as bridges can monitor the usage status and infrastructure security www finebook ir 20 WIRELESS SENSOR NETWORKS j ae y lt L Bellevue fo ATE eo 4 FIGURE 2 2 Traffic monitoring sensor network on Microsoft SensorWeb 22 CHARACTERISTICS OF SENSOR NETWORKS Wireless sensor networks are capable of observing the enviornment transferring data among their nodes and making decisions on the basis of these observations These networks are important for a number of applications most notably for target detection and localization surveillance and enviornmental monitoring Advances in miniaturization of microelectronic and mechanical structures have given rise to many battery powered sensor nodes that have sensing communication and process ing capabilities These sensor nodes can be network
233. ied hidden terminal problem As a result of this steps have to be taken to reduce or elim inate the occurrence of errors The collision avoidance algorithm presented below suggests a way to avoid such errors event void AMSend sendDone message t msg error t err if err FAIL call Leds ledOToggle 1 async event void AMA changed U event void RadioControl stopDone error t err www finebook ir 144 ALGORITHMS FOR WIRELESS SENSOR NETWORKS 85 COLLISION AVOIDANCE TOKEN BASED APPROACH 8 5 1 Token Based Approach The approach can be described very briefly using the following four abstract steps 1 Create a token and pass it to a node 2 If a node has the token it transmits the data that it wants to transmit 3 The node then transmits the token to its neighbor 4 Use this idea to count the number of nodes in the network The state diagram in Fig 8 6 captures the abstract behavior of each node in the token based approach for collision avoidance We implement this algorithm by building a component for it starting with its interface module CollisionAvoidance tokenBased uses interface Boot uses interface LowPowerListening interface Timer lt TMilli gt as Timer0 uses uses interface Send uses interface AMSend uses interface Receive uses interface PacketAcknowledgement FIGURE 8 6 State diagram for token based approach for collision avoidance www finebook ir Component Interf
234. iffusion with Simple MAC Assume that N sensor nodes are randomly placed in a grid of size MP Randomly few nodes send queries toward a region of interest The path taken by queries is decided by first sending interests We implemented an attribute list to define the type of interest or data message When a node receives an interest message it first checks wheather it has the property list of its neighbors The property list that the node maintains is the distance from the neighboring node to the final destination and the energy levels of the neighboring nodes If the node has this neighbor list it checks the last updated time of the neighbor list If this time is within the permissible limit this information Delivery Ratio uo ud co wc o eon Fk A 15 30 S0 100 Number of Queries FIGURE 14 7 Delivery ratio of sensor simulator www finebook ir EXPERIMENTAL RESULTS 267 is used to decide the next hop neighbor If the neighbor list does not exist or the last updated time is more than the desired time limit then beacon messages are sent out All the neighboring nodes receive this beacon message The neighboring nodes then send back beacon reply messages which update these properties in the neighbor list The nex hop neighbor decision is based on the GEAR protocol We give equal weight age to distance and energy factors After the query reaches the region of interest it is flooded to all the nodes in the region A visited node list is
235. ificantly longer to enable continuous collection of data Crau X V x 60 x 60 Node lifetime 5 X X 9 9 50 01 13 3 Evy Rx idle Lis Node lifetime for renewable _ recharge rate mAh energy routing nodes Fixed Tx cost mAh 13 4 These services share the resources of a node found among many fusionable ambient renewable MACS FARMS 1 deployed to route the data with the available software hardware and compatible radios as shown in Fig 13 3 In particular FARMS provides an innovative way to describe the reactive properties of the sensor network when sensing in a real time environment Evolution of the INSPIRE framework allows us to port distributed algorithms with a transparent power aware service meant for sensing applications www finebook ir 242 PERFORMANCE ANALYSIS OF POWER AWARE ALGORITHMS Tracking Sensing Sensing at Application Application Ultra low S Duty cycles Pt Pt Routing Neighborhood Assassin Data Robust 1 1 Sharing Bares Collection Dissemination SPEED 1 N Directed Diffusion Telos MicaZ Mica2 Dot Mica FIGURE 13 3 Comparison of traditional reactive and renewable programming models using compatible wireless radios 13 2 SERVICE ARCHITECTURE Stack architecture service architecture stack and service implementations and re lated details are shown in Figs 13 4 13 11 As in the previous work most of the simulation 3 has be
236. ilable in the src directory Simple Modules A simple functionality of each layer is defined here These are derived from base modules Users must implement their protocols at this level These are available in the samples directory The simple source code files for each layer are available in each subdirectory of the samples directory with respect to that layer The ned files for these simple source files are available in the src directory The ned files for new implementations are available in the subdirectories itself see item 4 in list preceding Base Layer Concept heading Conclusion We conducted various experiments to verify the data received by the subscribers from publishers This delivery ratio is 100 for smaller networks The delivery ratio decreases for larger networks according to directed diffusion www finebook ir REFERENCES 275 The user can run directed diffusion with 802 11 by including the class filename in the omnetpp ini file The topology number of queries number of nodes in the region the region size and the simulation time are configurable in the omnetpp ini File The user can also run various simulations using different seeds as described in the OMNeT 4 manual ACKNOWLEDGMENTS The authors acknowledge the assistance of students and faculty in the Sensor Network Group for their contributions toward the development of SensorSimulator The authors also acknowledge the CCT for providing infrastructure faci
237. ilures 0 EVB_LED1_ON else conPrintROMString Network crc Join FAILED Waiting rt then trying againW my_timer halGetMACTimer wait for 2 seconds while halMACTimerNowDelta my_timer lt MSECS TO MACTICKS 2 1000 routerState ROUTER STATE JOIN NETWORK j break case ROUTER STATE NORMAL check ping timeout if halMACTimerNowDelta my timer gt MSECS TO MACTICKS 1000 www finebook ir 227 228 INSPIRE reset timer my_timer halGetMACTimer long timeouts are done in segments of short intervals as the maximum timeout on the HAL MAC timer is platform dependent ping_cnt if ping_cnt PING_TIMEOUT ping cnt 0 send a ping routerState ROUTER STATE DO PING break case ROUTER STATE DO PING conPrintROMString Sending Ping n aplPingParent uses an APS ack in order to ensure that this node is still associated with the parent aplPingParent routerState ROUTER_STATE_WAIT_FOR_PING break case ROUTER STATE WAIT FOR PING if apsBusy break if aplGetStatus LRWPAN STATUS SUCCESS all is well failures 0 my_timer halGetMACTimer routerState ROUTERSTATE_NORMAL else failures conPrintROMString Ping failed n if failures MAX PING FAILURES failures 0 routerState ROUTER_STATE_REJOIN_NETWORK break case ROUTER_STATE_REJOIN_NETWORK A rejoin takes less
238. improved on by many individuals and was funded by several federal agencies The culmination of this work done by the LSU sensor network research group and others is recorded here Sensor programming is very different from traditional programming The authors have made an attempt to present the programming structure and implementation aspects of sensors under the framework of the nesC language Many researchers friends students and faculty contributed to this work We would like to acknowledge their contributions This work has been supported in part by the DoD DEPSCOR grant by the Office of Naval Research and by a PKSFI grant LEQSF 2007 12 ENH PKSFI PRS 03 from the Louisiana Board of Regents The authors would also like to thank the Indian Institute of Science for their travel support to the 100th Centenary Conference Countless people have provided assistance we have benefited from discussions with many of our colleagues around the country while working on this project We are grateful to them all and sorry that we cannot list all by name but a few need to be mentioned We would like to thank V Iyer for his several vital contributions to this work including working with LaTex and general proofreading We would like to thank Professor Ramamurthy and Professor M B Srinivas of the International Institute of Information Technology Hyderabad India for being co authors of many previous publications that helped us put together this book for FARM
239. in Table 4 8 Given n pendant vertices and their weights construction of a binary tree that minimizes the weighted pathlength is an interesting problem It has applications in decision tree and optimal code construction problems Some binary trees and their weighted pathlengths are shown in Fig 4 28 4 8 2 Spanning Trees Let G V E be a connected simple graph A cycle free connected subgraph T V with all the vertices of G is called a spanning tree A graph G and some of its spanning trees are shown in Figs 4 29a 4 29d Consider six cities shown in Fig 4 30a The distance between the cities are represented by the weight of the edges If a communication network has to be installed among the cities it is sufficient if we establish the link along the edges of the spanning tree shown in Fig 4 30b The cost of installation of the communication www finebook ir TREES 71 c FIGURE 4 26 a A complete binary tree b a full binary tree c an incomplete binary tree FIGURE 4 27 A tree with weighted pathlength 95 www finebook ir 72 DATA STRUCTURES FOR SENSOR COMPUTING TABLE 4 8 Binary Tree Weights and Level Numbers Pendant Vertex Weight Level liwi d 5 2 10 q 10 2 20 g 7 3 21 p 3 4 12 r 4 4 16 f 8 2 16 gt liwi 95 wires between two cities is proportional to the cost of the distance between them and hence the cost of the edge in the graph Thus the total cost of installation is proport
240. in its execution by calling the event function booted Inside booted we call the command function from the interface MyTimer start Periodic 200 which starts a clock that fires a pulse every 200 milli seconds Thus the control flow of execution that started with booted has jumped to start Periodic 200 As we have said before whenever a clock pulse is fired the event function ired from Timer is called and thus the control has flown now to the event functionMyTimer fired where we call the command function 1edO Toggle which toggles the red LED When the clock fires its next pulse after another 200 milli seconds the red LED toggles again This continues indefinitely as we have illustrated in Fig 7 2 We now can complete the application by specifying the components that provide the inter faces that we have used in the module TwinkleC and this is done in TwinkleAppC which shows the required wiring As we know LedsC provides the implementation for the interface Leds and MainC provides the interface Boot We should now also remember that TimeMilliC provides an implementation for the Timer interface Note that this has been renamed as SystemTimer0 configuration TwinkleAppC components MainC TwinkleC LedsC components new TimerMilliC as SystemTimer0 TwinkleC Boot gt MainC Boot TwinkleC Timer0 gt SystemTimer0 TwinkleC Leds gt LedsC 7 2 SENSING THE WORLD We will now look at one application wh
241. indicate that the packet was successfully sent See middle branch under the node labeled booted Also note that this part of the code appears in the event function Boot booted In the meantime if the node had received any packets the event function Receive receive is invoked where we turn the LEDs on setting the bits to ox7 to indicate that a packet has been received successfully module PacketSenderC uses interface Boot uses interface Leds uses interface Timer lt TMilli gt as Timer0 uses interface AMSend uses interface Receive uses interface SplitControl as PSControl uses interface Packet implementation uint16_t counter 0 message t pkt bool busy FALSE error t err void setLedBits uint8 t state void sendPacket RRR k k k k k k k Kk k k k k k k k k k k k k RR RRR KOR K RK RRR K KOR OR K K Bit mask corresponding to status of LEDs 0001 Booted 0011 Trying to send Packet 0101 Sent Packet 0111 Received Packet ROR k k k k k k k k k k k k k k k k k k k k k k k k k RR RRR k R KOR ke k k We have used a set of mask bits to choose which LEDs to turn on to signify the intermediate states of the PacketSender program Thus after booting we turn on the LED mask bits are 0001 then turn on ledO and led1 mask bits are 0011 to signify www finebook ir 128 SENSOR PROGRAMMING the fact that we are trying to send a packet and so on The timer is set to a predefined value TIMER FREQ module
242. ing light levels in the home environment The profile defines different device descriptions that belong to the profile including light sensor monochromatic switch remote control switching load controller and dimmer remote control A profile can consist of 216 device descriptors and can hold up to 256 clusters Each cluster can contain up to 216 attributes A device description contains a set of mandatory and optional input and output clusters from the profile Input clusters consist of attributes that can be set by other devices for example the light sensor has an attribute called ReportTime which controls the time interval between light readings Output clusters consist of attributes that supply data to other devices for instance the light sensor monochromatic LSM has one attribute in its output cluster named CurrentLevel which holds the current light sensor reading measured in 6 lux Mandatory clusters including every attribute within these must be implemented by the appropriate end devices Optional clusters may be implemented but if a device supports an optional cluster it must implement every attribute within that cluster 11 4 DISSEMINATION AND EVALUATION As this research is a multidisciplinary effort the programming of the sensor networks will be evaluated on the types of real time and energy constraint needs which have been addressed at the processing nodes and also how this stream of live data is been
243. ing portion of the code presented above the program does not increment the SendCount variable until the Read operation has completed and the conditional statement evaluated This program s semantics are very different in the split phase version of the code in which the Read command is called but does not halt the program flow When the reading operation is finalized the event ReadDone is signaled with the value of the reading and a message indicating whether the reading is a success or failure 5 2 2 Tasks Finally a task function represents the TinyOS notion of concurrency It allows a function call to be deferred for a later time when the scheduler can process less important tasks Multiple tasks can be queued up in the task queue which is a FIFO First in First out data structure Unlike commands and events a running task can be preempted by the system when an interrupt occurs www finebook ir 96 TINY OPERATING SYSTEM TinyOS Blocking Non Split Phase if Read SUCCESS SendCount NonBlocking Split Phase Start Phase Sensor Read Complete phase event void ReadDone error t err uint16 value if err SUCCESS 1 printf Value read by sensor is sa value j Syntax of a Task Function task return type task name gt lt input param gt task void computation void 5 3 EVENT DRIVEN PROGRAMMING Event driven programming or event based programming refers to a programming paradi
244. ing sensor networks and social networks 15 4 Design a sensor based tsunami warning system for tracking weather events in an unstructured enviornment 15 5 Describe how to use sensor networks to implement monitoring of a costal erosion problem 15 6 Apart from Monitoring activities list three other important applications of sensor networks www finebook ir 296 MATLAB SIMULATION OF AIRPORT BAGGAGE HANDLING SYSTEM REFERENCES 1 Anonymous Optimal data fusion in multiple sensor detection systems IEEE Trans Aerospace Electron Syst AES 22 98 101 1988 2 G Black and V Vyatkin Intelligent component based automation of baggage handling systems with IEC 61499 IEEE Trans Autom Sci Eng 6 2009 3 V T Le D Creighton and S Nahavandi Simulation based input loading condition optimi sation of airport baggage handling systems Proc IEEE Intelligent Transportation Systems Conf Seattle WA 2007 4 K Leone and R Liu The key design parameters of checked baggage security screening systems in airports J Air Transport Mgmt 11 69 78 2005 5 J C Rijsenbrij and J A Ottjes New developments in airport baggage handling systems Transport Plann Technol 30 417 430 2007 www finebook ir 16 Security in Sensor Networks If you spend more on coffee than on IT security you will be hacked What s more you deserve to be hacked White House Cybersecurity Advisor Richard Clarke 16 1 INTRODUCTI
245. ing the wiring details It is now easy to modify our initial implementation and interface built for PrintMessage to perform more complex tasks The interface PrintMessage multi shown below adds a new sending command send2 and its associated event send2 Done PrintMessage multi nc Implementing this new interface requires the addition of one new command in our PrintMessage implementation in which the event send2 Done will be signaled By now the role of events should be more apparent in interface declarations The notion of commands is fairly easy to understand Commands are functions that a component C calls in order to achieve some result Execution of a command might produce the desired result in addition to other exceptions What to do with the result and the exceptions is not the command executor s job rather it is the caller C s responsibility and what exactly C would like to do with the results and the exceptions should be defined in the component C In the current example we can imagine that the event send1 Done corresponds to the command sendl and the event send2 Done corresponds to the command send2 interface PrintMessage multi command error t sendl char msg 20 uint8 t len command error t send2 char msg 20 uint8 t len www finebook ir 104 PROGRAMMING IN NesC event void sendl Done char msg 20 error_t err event void send2 Done char msg 20 error t err 6 2 1 Tasks As we have seen before the com
246. ion for Real time Environment The bottleneck in writing code isn t in the writing of the code it s in understanding and conceptualising what needs to be done Shane Legg Wireless ad hoc sensor networks WASNs have drawn much attention in recent years from a diverse set of research communities Researchers have been concerned mostly with exploring application scenarios investing new routing and access control proto cols proposing new energy saving algorithmic techniques and developing hardware prototypes of sensor nodes This is described in the context of programmers and code base Typically traditional programmers use lots of inefficient techniques which need more computation As in the context of event driven programming it encapsulated as an even handler Making any unexpected loops to be present in the actual deployment Little attention has been devoted to actually measuring the lifetime of sensor net works Lifetime is defined as the useful time during which the sensor networks provide live datastreams before sensor faults occur or a percentage of sensors completely cease to function Since the normal sensor system s lifetime is in the order of many months to years especially in the case of ultra low duty cycling applications it is difficult to deploy and predict the lifetime of applications In this section we will use a software simulator that allows us to deploy a large sensor network This will enable us to accu rately measu
247. ional to the sum of the weights of the edges So we must find the spanning tree that has its sum of the weights of the edges minimum We define the weight of the spanning tree as the sum of the weights of its edges Given an undirected graph finding the minimum weight spanning tree is a very interesting problem The minimum weight spanning tree is also called the minimum cost spanning tree or shortest spanning tree Every connected graph has a spanning tree of G This tree T has n l edges These n 1 edges of the tree are called the branches The nonfree edges are called the chords If G is a disconnected graph Fig 4 31 it has more than one component In this case we can find one spanning subtree for each component and the collection of such spanning subtrees is called the spanning forest of G Figure 4 32 shows a spanning forest of the disconnected graph given in Fig 4 31 Let G be a graph with n vertices m edges and c components The spanning forest of G contains n c branches and there are m n c chords We define the rank of 10 5 7 8 a b c d FIGURE 4 28 Some binary trees and their weighted pathlengths www finebook ir TREES 73 2 I 8 5 6 7 3 4 b I 8 2 5 6 7 3 4 d FIGURE 4 29 a A graph G b spanning tree T c spanning tree T d spanning tree T3 the graph as the number of branches and the nullity of the graph as the number of chords For the graph shown in Fig 4 31a n 15
248. iques due to different data semantics and volumes to communication protocols due to various links and to security control due to different computational purposes of sensor nodes www finebook ir 24 WIRELESS SENSOR NETWORKS 2 3 NATURE OF DATA IN SENSOR NETWORKS The data produced by sensors in sensor networks have unique characteristics because of the goal of recording changes or rare events in many applications the nature of sensor nodes and the deployment of networks Streaming Sensor nodes produce data continuously A temperature sensor may produce data every second and can be represented as a tuple lt time sensor ID value gt This amounts to over 2 million tuples per month For a large network of sensors the total data will be in the order of terabytes per month too much to store This streaming nature raises several requirements for data processing Sensor data must be processed online and in network processing is necessary Further sensor data should be appropriately precomputed and stored in a convenient format for later queries Correlation Spatiotemporal correlation exists among sensor observations Because of the limited communication range of a sensor node a wireless sensor network requires a spatially dense deployment of sensors to achieve satisfactory coverage As a result the sensor observations about a single event from multiple sensors are spatially proximal and thus are spatially correlated Uncertainty The u
249. ir 236 INSPIRE CLUSTER HEAD SENSOR START LIFE TIME CH Establishes and request joining message HOP1 m Sensor in the neighbor responds to join CH adds it to its neighborhood list Sensor responds to beacon schedule and refreshes new reading HOP3 m CH aggregates new value from sensor HOP m MULTI HOP CH MULTI HOP CHy HOP N 20 MAX 80 MIN END LIFETIME FIGURE 12 4 Timing diagram for a hardware mote and processed If there is a threshold set for this particular process then an alarm will be enabled according to the current read value The on idle event can support the low power features of the target hardware enabling further energy savings Features of poll and on idle events are presented in Table 12 4 Both poll and on idle events are characterized 12 2 SOFTWARE MICROFRAMEWORK REQUIREMENTS The heart of the design is the event driven microframework which allows the building of robust reliable and resilient sensor motes Here the framework uses well defined TABLE 12 4 Sensor Processing Events Poll Initialize Process On idle Timeout Event alarm www finebook ir REFERENCES 237 object oriented techniques which are easy to maintain and also enables modeling state machines These machines are predictable and have deterministic state transition during their entire lifetime execution 12 2 1 State Machine I
250. it b p PARALLEL P1 wait for ack p p 1 mod 2 Receive receive no clock necessary P2 time out goto RESEND Receiver j Control bit b 0 repeat for ever receive packet p Receive receive is adequate to handle this situation bl extract control bit p d extract data p if bl b amp amp checksum valid p then send ack b b 1 mod 2 No ve ack is sent Go back and wait for the next transmission of the same data www finebook ir ALTERNATING BIT BASED ARQ PROTOCOLS 165 TRANSMITTER 4 alternating bit FSM Model Transmitter sO bit 1 c bit 1 frame getFrame transmit frame c bit is control bit sent by receiver go to s1 S1 receive if c bit bit bit bit 1 mod 2 frame bit packet transmit frame upon ack goto s1 Pseudo code for nesC global bit 1 c bit 1 booted frame getFrame transmit frame receive actype extract acktype if acktype ve then frame getFrame transmit frame else transmit frame Receiver 4 alternating bit S0 control bit 0 goto s1 S1 receive bl extract control bit p d extract data p if bl b amp checksum valid p then send ack b b 1 mod 2 No ve ack is sent 9 6 1 A Generalized Version of the Previous Protocol We can generalize the above protocol presented above by going back by n trans
251. iveMessageC components AMSenderC components Timer lt TMilli gt as Timer0 AppC Boot MainC AppC Packet AMSenderC AppC AMSend gt AMSenderC AppC Send gt AMSenderC AppC Receive AMReceiverC AppC TimerO gt Timer0 Pseudocode Implementation event void Boot booted mode initialize le EU Timer0 startPeriodic event void TimerO fired if mode transmit amp counter 0 AMsend to node B data C GELO www finebook ir 148 ALGORITHMS FOR WIRELESS SENSOR NETWORKS if mode transmit amp counter 10 1 AMsend to node C data emu 41 10 event void Receive if mode receive data packet J States of Transmit The concept described above can be used to avoid collisions as we discuss below Divide time into slots Insert permitted transitions at each slot For example the time axis is divided into intervals of period T Each period T is divided into n slots Thus T slot 1 slot 2 slot n e Indicate the possible transmissions in each slot For example slot 1 A B C D slot 2 E F slot 3 G H I J etc Assign to each node The list of interfaces we used in developing this application component are shown in the program above along with the components that we selected for their imple mentation Schedule Based Transmission In order to implement the schedule we use a small counter that tells us when to
252. knowledge of graphs greatly simplifies the task of sensor programming 4 6 1 Network Localization Sensors are typically deployed in an area to measure spatial and temporal characteris tics of events of particular interest This implies that these sensors need to convey data FIGURE 4 9 Three linked lists www finebook ir 58 DATA STRUCTURES FOR SENSOR COMPUTING ge ET EB HN M M FIGURE 4 10 a A list b a linked list about the event and also to relay information about the location of where that event occurred before any useful task can be done with that data In network localization knowledge of graphs can be useful in certain scenarios to provide an estimate of the distance between nodes in a network Several node localization algorithms have been published that make use of concepts in graph theory 4 6 2 Data Aggregation The ability to aggregate data from distant nodes for more efficient transmission is an important data processing primitive for sensor networks particularly since it results in huge energy savings extending the life of a deployed sensor network For data aggregation to be successful routing algorithms have to make intelligent decisions when choosing nodes to partake in data transmission from the source node to the data sink or otherwise risk depleting the energy of nodes in most frequent use The most natural data structure that comes to mind when routing information based on associated costs and other para
253. l by formulating a deterministic key predistribution scheme proposed by the KKID algorithm 3 that is based on assigning keys to sensors by placing them on a grid 16 2 Develop a programming tool by using a cryptographic authentication mecha nism by attempting to add denial of service resistance to existing protocols www finebook ir 304 SECURITY IN SENSOR NETWORKS 16 3 What are the programming constraints in developing a secure sensor network to be deployed in a hostile environment that is prone to malicous attacks 164 Why is effective collision detection problematic in wireless networks 16 5 Give an example of a problem algorithm that could be used as a client puzzle and implement it in a program REFERENCES 1 J Stankovic A Perrig and D Wagner Security in wireless sensor networks Commun ACM 47 53 57 2004 2 H Chan A Perrig and D Song Random key predistribution schemes for sensor networks Proc IEEE Security and Privacy Symp 2003 May 2003 3 R Kalindi R Kannan S S Iyengar and A Durresi Sub grid based key vector assign ment A key pre distribution scheme for distributed sensor networks J Pervasive Comput Commun 2 1 35 43 March 2006 4 C Karlof and D Wagner Secure routing in wireless sensor networks Attacks and coun termeasures Proc 1st Int IEEE Workshop on Sensor Network Protocols and Applications Univ California Berkeley 2003 5 A Perrig R Szewczyk V Wen D Culler
254. l resource usage being of no consequence energy can be replenished for most devices communication between multiple devices is very cheap and reliable It is in this respect that wireless sensor networks chiefly differ from conventional networks With nonrenewable sources of power and very little on board power computation and communication come at a very high price As wireless sensors become more pervasive because of their lower cost we can expect the pro liferation of sensors to exceed those of traditional computing devices Hence there is a need to develop new energy aware routing algorithms and aggressive power Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright 2011 John Wiley amp Sons Inc 131 www finebook ir 132 ALGORITHMS FOR WIRELESS SENSOR NETWORKS management schemes for this newly emergent class of computing devices 4 In this chapter we will discuss several concepts challenges properties and algorithms unique to wireless sensor networks such as Communication patterns prevalent in sensor networks Physical components of sensor nodes Properties of wireless sensor networks Networking layers Routing Also we will show how some of these concepts can be implemented using nesC code and pseudocode 81 STRUCTURAL CHARACTERISTICS OF SENSOR NODES The term sensor nodes also called motes r
255. l wireless sensor network consists of spatially distributed sensors that co operatively monitor some physical phenomena This network formed by the sensors could contain nodes with varying capabilities and sometimes completely different underlying platforms which must collaborate and communicate with each other There are several advanced research platforms for wireless sensor networks in use by researchers globally with each platform offering unique differentiators such as sensor size power consumption nature of operating system or basic sensing abili ties 10 9 In this chapter we examine some sensor platforms and their associated software tools commonly in use today Most tools used in the management of wireless sensor networks can be classified into three distinct categories see also Fig 3 1 Sensor level Server level Client level 3 1 SENSOR LEVEL This level consists of devices that measure some physical phenomena or quantity such as sound motion light intensity or temperature and convert it into some quantifiable form that can then be read by devices or human observers Each sensor s mote contains an onboard sensing communication power and processing module that allows it to perform sensing tasks In the following sections we discuss some of the most commonly used sensor platforms and their features that make them unique 3 1 4 The Mica Family The Mica family of sensors is one of the most common sensing platforms in
256. lanar if and only if it doesn t have a Ks or K5 as its subgraph Planarity testing is widely used in wireless sensor networks for problems as sociated with coverage localization routing topology control and detection of holes 4 6 5 Graph Coloring Concepts in MAC Layer Protocols The minimal graph coloring problem is defined as the minimum number of dif ferent colors that are required to color the vertices of a graph such that no two adjacent vertices share the same color This fundamental graph theoretic problem is widely used in sensor networks for various applications Ranging from design ing efficient interference free MAC layer protocols service discovery data aggre gation Fig 4 13 TDMA slot assignments and other scheduling assignments are approached using the graph coloring problem For example in scheduling for wire less sensor networks the individual nodes must be scheduled in such a manner that FIGURE 4 12 A simple illustration of node localization www finebook ir 60 DATA STRUCTURES FOR SENSOR COMPUTING FIGURE 4 13 A simple illustration of data aggregation nearby nodes do not attempt to transmit data at the same time or should be scheduled in such a way that the nearby nodes use different frequency channels for trans mitting to avoid cochannel interference By efficiently scheduling the nodes using graph coloring one can minimize the energy consumption and latency in a given topology 4 6 6 Isomorphism
257. language that is simple to use and understand Function www finebook ir MANA 2I4 U92 X1o0 oeure1 Sururure130Jd 1osuos pojnquysip e jo urouoxe V p H201SLI uoneJous8 Sjueuuoui ua 9 sisAfeue opo Snqep pue quowdojaneq JUIOdMAIA 8u S pejueuo palga rwyposy MIA 2I qU92 uone idde IOnuoo 66v19 DAl IEI Dal quiodmel asensue SuriuouupuAs uongeyueumo oq queuJesSeue eq K Suipuer uonde x uoneiueuue duu uon unj suonexiunululo sam ume Chung sonsouseiq C Suns l uiuluog v 49 www finebook ir 50 DATA STRUCTURES FOR SENSOR COMPUTING blocks are modular software units can contain internal states and represent the inputs and outputs of the function Libraries of common blocks are provided for building control programs General purpose languages such as FORTRAN or C are employed to specify the processing in sensors More recently object oriented languages have been used to program controllers Domain specific languages with extensions to specify data fusion functions tailored to the needs of particular ap plications are likely to be useful Development and debugging environments for a DSN should support modular independent programming of different sensors Key distributed programming constructs can be provided to the programmer by distributed system services or they can be embedded in the language and implemented
258. late communication networks and other distributed systems The OMNeT model is a collection of hierarchically nested modules as shown in Fig 14 1 The top level module is also called the system module or network This module contains one or more submodules each of which could contain other submodules The modules can be nested to any depth and hence it is possible to capture complex system models in OMNeT Modules are distinguished as being either simple or compound A simple module is associated with a C file that supplies the desired behaviors that encapsulate algorithms Simple modules form the lowest level of the module hierarchy Users implement simple modules in C using the OMNeT simula tion class library Compound modules are aggregates of simple modules and are not directly associated with a C file that supplies behaviors Modules communicate by exchanging messages Each message may be a complex data structure Messages may be exchanged directly between simple modules on the basis of their unique ID System Module CM Compound Module Messages between Simple Modules SM Simple Module v v FIGURE 14 1 Simple and compound modules in OMNeT www finebook ir SIMULATION DESIGN 257 or via a series of gates and connections Messages represent frames or packets in a computer network The local simulation time advances when a module receives messages from another module or from itself Self messages are used
259. layers physical datalink network trans port and application and three planes energy management mobility management and task management as shown in Fig 2 3 In this model a stack would be main tained across multiple nodes on the network with sinks clusterheads elected leaders implementing the lions share of the stack The physical layer deals with A D D A conversion modulation demodulation and transciever techniques such as RF carrying The datalink layer is concerned mainly with media access control and error recovery The network layer deals with routing because of the one to many and many to one data transmissions that are so common in a WSN standard routing protocols are not very useful here On top of this network protocols seldom consider power usage and its effect on network lifetime both critical factors in WSN design Because of power and processing limitations WSN cannot afford to implement standard routing tables and generally must adopt some lightweight alternative The transport layer provides ports or transport interfaces to various applications just as in the ISO OSI model Finally the application layer decomposes and implements application specific tasks by utilizing the lower layers www finebook ir 22 WIRELESS SENSOR NETWORKS Task Management Plane Mobility Management Plane Energy Management Plane Application Layer Transport Layer Network and Routing Layer Datalink MAC Layer Physical Layer
260. lementation components TrackingC as App components MainC components new TimerMilliC as TimerO components new TelosbSensorC App Boot gt MainC App Timer0 gt Timer0 App ReadStream gt TelosbSensorC Read Stream is needed to read the temperature from the environment and timers are needed for sampling We first read the initial values to store them in a vector Later during tracking we sample and read the sensor values and compare the new values with the values in the vector and look for significant changes in the values The implementations of interfaces are shown below We have chosen the ReadStream implementation for the TelosB mote Using these implementations of the interfaces we implement our component as follows First We start the two timers one single shot and the other periodic When the single shot fires we sense the initial temperature and store it as the reference value The periodic clock is used to sample the environment periodically and check whether the temperature has changed significantly implementation define threshold 0 5 int templ temp2 www finebook ir QUERYING IN RUMOR ROUTING 189 FIGURE 9 4 Querying in rumor routing event void Boot booted initialize two timers as TimerO startOneShot 100 and Timerl startPeriodic 2000 event void Timer0 fired templ Temperature read event void Timerl fired temp2 Temperature read if abs templ temp2
261. liability in large wireless sensor networks Proc IEEE Int Conf Sensor Networks Ubiq uitous and Trustworthy Computing 2008 pp 480 485 6 J Polastre J Hill and D Culler Versatile low power media access for wireless sensor networks Proc 2nd Int Conf Embedded Networked Sensor Systems SenSys 04 ACM New York 2004 pp 95 107 7 M G Schultz E Eskin et al Data mining methods for detection of new malicious executables Proc IEEE Symp Security and Privacy May 2001 8 W Ye F Silva and J Heidemann Ultra low duty cycle MAC with scheduled channel polling Proc 4th Int Conf Embedded Networked Sensor Systems SenSys 06 ACM New York 2006 pp 321 334 www finebook ir 13 Performance Analysis of Power Aware Algorithms Efficiency and quality are of equal importance Both come from experience not from study Study as you go don t assume that you re ready for the real world because you studied first Jon Davis 13 1 INTRODUCTION In the previous chapter most of the application development framework was been discussed for specific deployment such as the IEEE 802 15 4 ZigBee Alliance These frameworks make application development and deployment very transparent to developers that hide some of the unique performance characterstics of sensor networks In this chapter we will use some of the parameters that are needed for scalable applications such as node counts in the neighborhood of 100 and power dissi
262. lient networks by using a mesh routing architecture Other functions of the network layer are to enable proper use of the underlying MAC layer by providing a suitable interface to other upper layers The routing protocol in use in a ZigBee network chooses the lowest cost energywise route to the intended destination The following factor is used total collaborative resource needed Global resource available in individual sensor This factor is calculated on routing or completing a sensor network task without having a single failure keeping cross layer energy consumption to a minimum 11 2 3 Datalink Layer The low level layers deal with radios and a suitable MAC that allows for communi cation with its neighbors without any central coordination The datalink layer is also responsible for aggregating data from all its neighbors and sending an acknowledg ment when it receives sufficiently reliable data to higher network layers 11 2 4 Physical Layer This layer consists of the actual radio and is IEEE 802 15 4 compliant 11 3 ZigBee APPLICATION DEVELOPMENT ZigBee basically uses digital radios to allow devices to communicate with one another A typical ZigBee network consists of several types of devices A network coordinator is a device that sets up the network is aware of all the nodes within its network and manages the information about each node as well as the information that is being transmitted received within
263. lities for the development of the Sensor Network Laboratory PROBLEMS 14 1 Provide a two page essay on how much realism can be achieved in a network simulator 14 2 Compare the features of J Sim GloMoSim and OMNet 4 14 3 Why are traditional methods of network simulation inadequate for sensor network simulation 14 4 Provide a brief write up describing the distinguishing features of the network simulator discussed in this chapter REFERENCES l http www cs rpi edu cheng3 sense 2 http www isi edu nsnam ns ns documentation html 3 Anonymous Information Technology Telecommunication and Information Exchange be tween Systems Local and Metropolitan Area Networks Specific Requirements Part 11 Wireless LAN Medium Access Control MAC and Physical Layer PHY Specifications Technical Report IEEE Standard 1999 4 J Agre L Clare and S Sastry A taxonomy for distributed real time control systems Adv Comput 49 303 352 1999 5 I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci Wireless sensor networks A survey Comput Networks 38 393 422 2002 6 S Basavaraju Sensim A Wireless Sensor Network Simulation Template M S Project Dept Computer Science Louisiana State University Baton Rouge 7 V Bharghavan A Demers S Shenker and L Zhang Macaw Amedia access protocol for wireless lans NACM SIGCOMM 1994 1994 8 K Fall and K Varadhan Ns 2 Network Simulator Technical Report Univ C
264. ll i If a list is implemented by keeping the list atoms in an array then we have a linear list A typical linear list representation is in the form of an array A 0 N where A 0 holds the number of items in the list Such a representation requires limiting the number of entries to N and is clearly wasteful in the sense that A A 0 1 A N are unused To keep a collection of several linear lists is awkward Say for example that the lists A B C D M N O and W X Y Z are to be kept and the lists are allowed to grow to hold as many as 10 entries each The lists may be regarded as a two dimensional array A 0 3 0 10 Table 4 3 describes the data structure of the lists described above Now consider holding all three lists in one array with an array of pointers into the array shown in Table 4 4 TABLE 4 4 Array of Three Linked Lists Position Data Link 1 A 5 2 X 11 3 M 8 4 W 2 5 B 9 6 O 0 7 D 0 8 N 6 9 C 7 10 Z 0 11 Y 10 www finebook ir LINKED LIST 55 Event Pool FIGURE 4 7 An event pool The array of pointers is called an index This has the advantage of removing the upper bound on the list length the only requirement is that the lists do not have a total length greater than the array size Each list occupies only the space it needs and new lists may be easily added at the end see Fig 4 7 The problem here is however in inserting an item into the lists The room to insert E in list 1 to make A B C
265. llision and maintenance of messages Message Overhead Message overhead complexity for both synchronous and asyn chronous operations is Messages 2 x E x number of message types 4 5 www finebook ir 90 DATA STRUCTURES FOR SENSOR COMPUTING where E is the number of edges in the network This is similar for both types of network For the synchronous protocol model time is measured as Messages O D 4 6 where D is the diameter For a synchronous model Messages O N 4 7 where N is the total number of nodes in the network In the worst case scenario an async protocol may construct a chain The synchronous network is more optimal as the diameter is always less than or equal to the total number of nodes PROBLEMS 4 1 What are some of the properties of data structures that enhance sensor pro gramming in the context of distributed sensor networks 4 2 Define a few efficient data structures for spanning tree protocols 4 3 Explain in your own words the concept of homeomorphism 4 4 Illustrate using examples the concept of isomorphism 4 5 What is the fundamental difference between a graph and a tree 4 6 Consider the abstract data type stack Specify clearly the interface operations for this data type Make sure that you have accounted for all the exceptions that may occur 4 7 Repeat Problem 4 1 for the data type tree Choose any type of tree 4 8 Consider two sensor nodes located adjacent to each other We would like to
266. lopment in sensor networks has been left mostly untouched We present this book to focus on software development in sensor networks This book provides the basics needed to develop sensor network software and supplements it with many case studies covering network applications We also examine how to develop onboard applications on individual sensors how to interconnect these sensors and how to form networks of sensors although the major aim of this book is to provide foundational principles of developing sensor networking software and to critically examine sensor network applications Courses in sensor networks do not provide direct access to sensor networking equipment and software However the need for hands on experience is essential for a high quality education We have structured the examples in the book in such a way that a teacher can demonstrate the material on a small network of four or more sensor nodes and a laptop all of which can be hand carried to any room and set up on a teacher s desk It is our hope to impart to students and by extension any reader of this book an understanding of sensor network programming based on general principles with an aim to develop actual software Thus in Part II we provide xiii www finebook ir xiv PREFACB implementations of various algorithms ranging from very simple to more elaborate in complexity Features This is a practical book The book contains many figures pseudocode and actual
267. ltiple times for academic distinction and most recently nominated as the most outstanding computer science senior During his 4 years at Louisiana Tech he held various leadership positions including the most coveted vice president of the local ACM chapter president of the Robotics Club and Louisiana Tech Program ming Team Czar His research interests include high performance computing HPC machine learning and wireless sensor networks Having done exemplary work in the design and implementation of high availability tools for clusters he was charged with the development of a high availability solution for the renowned Louisiana Optical Network Initiative LONI After being accepted in the Google Summer of Code Program under the mentorship of researchers from Oak Ridge National Laboratory he greatly enhanced the widely popular OSCAR software stack for HPC clusters Currently he works as an undergraduate student researcher for the Center for Secure Cyberspace under the supervision and mentorship of the well accomplished Dr Vir V Phoha in the Sensors and Machine Learning Group www finebook ir Notations and Abbreviations Clarity of presentation is critical to effectively communicate knowledge 5S S Iyengar ACRONYMS AG acquaintance group API application program interface ARQ automatic repeat request ATR atomic target recognition BC block cut e g block cut tree BFS DFS breadth first depth first search BHS baggage han
268. m node A packet id get packet id packet p data get data packet p if packet id i buffer i data i i 1 send ve ack else send ve ack i Transmitter Selective Repeat Selective Reject S0 k 0 sl while k lt 7 send a k goto s2 s2 receive Cc control bit if ack for c is ve then send a c Receiver Selective Repeat Selective Reject S0 i 0 while i lt 7 receive packet p from node A packet id get packet id packet p data get data packet p if packet id i buffer i data l s X od send ve ack else send ve ack i www finebook ir 170 TECHNIQUES FOR PROTOCOL PROGRAMMING nesC code booted i O receive In some applications in the real world it is more convenient to refer to the nodes using their network wide unique address than according to their location or data In the following paragraphs we briefiy discuss the need for address management in wireless sensor networks 9 8 NAMING AND ADDRESSING A sensor network is fragile and the sensors can malfunction at any time Thus the nodes in a WSN must have the capability to allocate unique addresses to each individual node and manage these addresses when new situations arise Consider for example a node which acts as a bridge between two subnetworks this node fails for some reason the two subnetworks will lose connectivity and for a n
269. me to establish the network path Along that path are numerous router nodes and clusterheads Each of these nodes helps manage wireless traffic to piggyback the data to the destination creating a multihop page between the data source and the destination node When a new data update is requested all of its neighboring nodes participate to enable the data to travel separately through this maze of the networked nodes Each hop along the way is an opportunity to introduce power savings and thus add to optimizing the overall power consumption of the sensor network While manyfactors contribute to energy wastage unscheduled overhearing and channel polling represent the largest and fastest growing portion of total energy wastage In fact studies 1 have shown that 8046 of the time nodes in a dense sensor network are affected by local ambient conditions such as collision overhearing and control packets The example shown in Fig 13 2 illustrates the harmful effect of even a small degree of energy wastage during idling under ideal conditions When lifetime decreases such as during transmittance or when packet loss rates increase performance can significantly deteriorate In fact studies have shown that it is more efficient for a routing algorithm to use a minimal number of nodes such as cluster heads or multihop leader nodes to schedule its data delivery One way to avoid the negative effects of idle wastage x number of nodes on life time
270. media access layer multiagent system multipoint control unit microelectromechanical system maximal outer planar graph objective modular network testbed product lifecycle or public limited company publish subscribe group quality of service random access memory read only memory reduced function device radiofrequency identification tag as in e g airport security random structures and algorithms reallly simple syndication run to completion real time OS operating system ready or request to send scheduled channel polling sensing health with intelligence modularity mobility and experimental reusability short interframe space security management system systems planning engineering and evaluation device Transmission Control Protocol Internet Protocol TinyOS simulator Unified Modeling Language Universal Serial Bus Universal Time Coordinated wireless ad hoc sensor network wireless sensor network www finebook ir PARTI Overview www finebook ir 1 Introduction The creation of genuinely new software has far more in common with developing a new theory of physics than it does with producing cars or watches on an assembly line T Bollinger Software that drives the operations of sensors and communication among sensors is basic to any meaningful application of sensor networks The goal of this book is to provide an understanding of how this software functions how it allows the sensors to gat
271. meters is a connected graph It is for this reason that several tree based data aggregation algorithms exist such as the collection tree protocol CTP in TinyOS Multiple Producers Single Consumer Timer Tasks o Events Events FIGURE 4 11 Circular linked list doubly linked list www finebook ir IMPORTANCE OF GRAPH CONCEPTS IN SENSOR PROGRAMMING 59 4 6 3 Collaborative Processing In most sensing tasks sensors rarely rely solely on their own sensing abilities but rather distribute a sensing task among a subgroup of sensors and collaborate on that task One example that comes to mind is any tracking task Most entities being tracked are mobile in nature and for this reason relevant sensors around the region of interest have to be selected dynamically for collaboration with other sensors on the basis of location and other factors The task of choosing an appropriate sensor can be greatly simplified by applying concepts from graphs 4 6 4 Planarity Testing In graphs the edges between nodes may intersect This could be due to either of two factors 1 the graph is nonplanar or 2 a planar graph is not drawn properly to exhibit its planar structure To ensure that a given graph is planar a well known computer science problem several well known practical algorithms have been proposed in the literature that run in O n time where n is the number of vertices One of them is Kuratowski s theorem which states that a finite graph is p
272. mission steps and starting to retransmit them We will illustrate this with an example below www finebook ir 166 TECHNIQUES FOR PROTOCOL PROGRAMMING 9 6 2 Example Let N e4 Let us use a buffer to store the packets To start with fill the buffer with eight packets prepending each packet with a control byte so that the content of the buffer buff will be buff 0 po 1 p 2p2 3p3 4p3 5p4 6ps p6 Start sending each packet from the left onward 0 po 1 p1 2 p2 one at a time while simultaneously receiving the acknowledgments and processing them Let b be the control byte of the last packet that was sent The recipient is assumed to send a positive acknowledgment ACK whenever the packet has been received correctly Negative acknowledgments indicating that the packet was not received successfully are not sent by the recipient When the sender receives a ACK it first extracts the control byte say 0 from ACK It then can conclude that the packet 0 po has been successfully received Suppose that it next receives a ACK for which the control byte is 4 It then concludes that the packets 1 p 2 p2 and 3 p3 have not been received correctly It thus sets the value of b to 1 and the packets are sent starting from the new value of b rather than from it sold value The code below shows the generalized version of the ARQ protocol Because of the multiple threads of reasoning involved the control flow in the program is somewhat invol
273. modules prewired configurations and predefined types supported in nesC In order to acquaint ourselves with some of the simple interfaces modules and configurations let us begin with an example provided in the TinyOS package called PowerUp This program uses two interfaces Boot and Leds which are already provided in the system and does not seem to require any other interfaces PowerupC nc module PowerupC uses inter face Boot uses interface Leds implementation event void Boot booted call Leds ledoOn j PowerupAppC configuration PowerupAppC implementation components MainC components LedsC components PowerupC The following statements denote the fact that the interfaces Boot and Leds used in PowerupC are provided implemented in the components MainC and LedsC respectively provided in the system PowerupC Boot gt MainC PowerupC Leds gt LedsC Thus note that there are no interface declarations in this example We are already familiar with the Boot interface What is more interesting is the interface Leds which specifies commands for controlling the LEDs light emitting diodes mounted on the motes by the manufacturer The Telosb mote supplied byCrossbowTechnology supports three LEDs of colors red green and blue 1 The interface Leds see code below provides commands for controlling each LED Additionally it provides commands for obtaining and setting the mask bits that are us
274. mple simulates the behavior of the application layer This module com municates with the NetLayerBase Module through gates to schedule any messages New applications can be incorporated to this module The functionality of this module is described in greater detail for the directed diffusion implementation The simple network module simulates the packets sent and received by the nodes in the network The network module initially receives application layer messages from the AppLayer module and adds the network header to it The particular features of this layer depend on the protocol implementation Directed diffusion with GEAR is implemented at the network layer as described in the next section The packet structure of the network layer sent to the MAC layer has the next hop in the route The MAC layer provides the interface between the physical layer and the routing layer It has the basic functionality of media access the functionality of this module is described in greater detail for the 802 11b implementation Such a modular structure of entities simulated with OMNeT 4 makes our simu lation more flexible than ns2 14 5 IMPLEMENTATION DETAILS Using the above mentioned design for the simulator we implemented directed diffu sion at the network layer and compared the performance with the existing simulator ns2 MAC 802 11b is also implemented at the MAC Layer and the performance is compared with ns2 with directed diffusion at the network laye
275. mprove the reliability of sensing tasks It is possible to densely deploy sensors because of their small size and low cost If a sensor node fails other sensors close to it can accomplish its task by working collaboratively This is especially useful in environments where it is impossible to replace sensors Sensors working collaboratively can also benefit resource conservation in sensor networks Sensors can be networked through physical connections or wirelessly In early sensor systems few sensors were deployed closely such as in industrial control or medical monitoring and were connected through wires to a control unit With low volume low cost and wireless communication the sensors can be deployed in a large area Even in densely deployed systems wireless communication reduces the complexity of wiring Thus wireless sensor networks have emerged as a new information gathering paradigm with large numbers of sensors collaborating We next describe sensor network applications briefly followed by the characteristics of sensor networks As our focus is on data processing issues the nature of sensor data is examined in this section 2 1 SENSOR NETWORK APPLICATIONS Sensor networks are deployed for collecting information on entities of interest The availability of low cost low power and multifunctional intelligent sensors enables military applications such as battlefield surveillance nuclear biological and chemical attack detection and ci
276. n infinite region A maximal outer planar MOP graph is an outer planar graph in which the addition of an edge between any pair of nonadjacent vertices will resultina non outer planar graph A MOP graph is Hamiltonian Several applications of the concepts discussed in this section exist in sensor net works Some examples of these include the location of holes in sensor network coverage and optimum deployment of sensors with the lowest number of nodes 4 12 COLORING AND INDEPENDENCE By coloring a graph we mean assigning colors to each vertex of the graph Some authors have studied the problem of coloring the edges of the graph They call it edge coloring Coloring the vertices is said to be vertex coloring We restrict our discussion to vertex coloring and so using the term coloring we mean only vertex coloring A perfect coloring is to assign colors to each of the vertices such that any two adjacent vertices get different colors Figure 4 45 shows a graph G and three different perfect colorings for G The main interest here is to color the graph using a minimum number of colors Such a coloring is called a perfect minimum coloring and the number of colors used for minimum coloring is called the chromatic number www finebook ir 84 DATA STRUCTURES FOR SENSOR COMPUTING w Y W White R B Blue G R Red Y Yellow BK Black BN Brown V Violet P Pink FIGURE 4 45 A graph with three perfect colorings The chromatic
277. n data structures will yield very efficient algorithms for several important classes of problem Because of the crucial importance of sensor data organization this chapter deals with various data structures used in the design and analysis of sensor based algorithms One of the basic facts that we need to understand is that sensor programming is different from traditional programming The high level coding in sensor programming is typical for any active object oriented framework Furthermore the design is layered with a tiny operating system the foundation for preemptive and roundrobin kernel and other basic services such as event driven and CPU pools Figure 4 1 is essentially a design based schematic diagram of program flow Figure 4 1a is a schematic flowchart of a quickstart application while Fig 4 1b is a flowchart of an event driven sensor application running on top of a cooperative vanilla kernel At the highest level the flowcharts are similar in that they both consist of a main loop surrounding various programming constructs But the internal structure of the main loop is very different in the two cases As indicated by the heavy lines in the flowcharts the quickstart application spends most of its time in the tight event loops designed to busy wait for receiving events such as communication update events In contrast the event driven application spends most of its time right in the main loop The wireless sensor frame
278. n the graph can be considered as an induced subgraph of the graph with one vertex 3 Any edge u v can be considered as an induced subgraph induced by u v 4 The relation subgraph of is reflexive antisymmetric and transitive Unlike simulated conditions of sensor deployments which typically assume ideal conditions for sensors real world deployments are usually faced with a barrage of issues stemming from unanticipated scenarios such as the effects of the environment landscape on sensor communication or even interference by other RF emitting de vices To handle such issues communication routines in sensor networks must possess dynamic capabilities computing alternative communication routes when disruptions occur and as such rely on critical concepts discussed in the following sections Sections 4 7 5 and 4 7 6 introduce the topics of homeomorphism and isomorphism in terms of localization and routing in sensor networks 4 7 55 Homeomorphism Let G be a graph Let vi v2 vj be a path and let the degree of U be 2 in G The pairs vi v2 and v2 v3 are called series edges The operation of replacing the two edges vi v2 and v2 v3 by a single edge vi v3 and removing the vertex V is called the merging of series edges If u v is an edge the operation of introducing a new vertex w and making it adjacent to both u and v and then removing the original edge u v is called insertion of a vertex of degree 2 Two graphs G an
279. nd RLINK for the right link and every list has pointers to the left end and right end Thus a typical doubly linked list resembles the structure shown in Fig 4 10b Note that there is one linked list found by following RLINK to a node and another list can be found by following LLINK to a node It is particularly easy to delete a node in a linked list The links allow you to find all different nodes Suppose that a node X has left and right adjacent nodes FIGURE 4 8 Continuous list www finebook ir IMPORTANCE OF GRAPH CONCEPTS IN SENSOR PROGRAMMING 57 TABLE 4 7 Binary Search Example ABCDEMNOVW X Y Z 123 45 67 8 9 10 11 12 Node X may be deleted by applying the following instructions RLINK LLINK X gt RLINK X LLINK RLINK X LLINK X This is seen to produce the following doubly linked list structure which effectively bypasses X The two lists in the doubly linked list may themselves be circular see Fig 4 11 To insert a node containing a desired data item we first must acquire an unused node For this purpose we keep a special linked list called the free list which holds nodes that are not currently used Such a list may be initialized when memory is first allocated for nodes 4 6 IMPORTANCE OF GRAPH CONCEPTS IN SENSOR PROGRAMMING Knowledge and understanding of the fundamental properties of graphs is critical in distributed sensor programming In this section we examine some of the ways in which
280. nes the minimal set of abstraction prim itives that are based on distributed WSN runtime conditions that make it transparent to the programmer Given the increasing importance of optimal quality of service in en hancing the lifetime of sensor networks the converse problem of energy equivalence in routing is equally significant and not yet addressed in the literature in a quantifiable manner We address both of these issues QoS by modeling a fixed lifetime resource model related work and a renewable energy model In particular we extend the common QoS parameters which are uniquely interdependent in both the models to TABLE 13 2 Wireless Communication Dependences Data Reception Data Transmission Neighbor Management Message arrives from link Abstracted link control Cooperatively managed parameters Service dispatches Abstracted link feedback data Service mediates interaction using table Network protocols establish References to packets No policy on admission associated with this message eviction by HP Naming addressing Link Power Scheduling information Filtering www finebook ir peurquioo s1oAe OVIA pu YIOMIOU 10 UMOPYLIIG pe uI Ao JSV 1osu s p 3oKeT OVIN ut Vp Jo uonsuen Jamo Surwous uoneordde SWAVA 9 tTo 2uoo uoneordde WIM speourojsnpo se sopou 07 YIM eoueurog1od OVW q SurpreA10j ewp WIM ooueuuojgred JY v seumogi e qeeuar pue eurtou 10g uosnrduioo soueuojied OVW ZTEI TANIA p ZL api LE
281. networks in agriculture They investigated sensor network design while considering www finebook ir 18 WIRELESS SENSOR NETWORKS FIGURE 2 1 On Great Duck Island in Maine wireless sensors at 1 and 2 pass data to a gateway node shown at 3 It is then passed to the base station at 4 and potentially sent over the Internet via the satellite dish at 5 the structure of work activities in agricultural production environments As a result some activities can be completed much more efficiently For example a vineyard can be sprayed only in places where there is a risk of powdery mildew Grape Networks has transferred such research achievements into commercial products Using Internet and mobile wireless mesh sensor networks farmers can monitor and receive alerts with a PC PDA or mobile phone on the environmental information such as soil moisture microclimates and diseases in vineyards or open fields 2 1 2 Sensor Networks Developments in geographical information system GIS satellite remote sensing and global position satellite system GPS technologies helps people receive disaster alert information more rapidly and accurately These techniques have disadvantages such as low resolution and weather interference because of long distances With the use of small and low cost sensors a large number of sensors can be spread in situ Various environmental data can be collected in real time and monitoring resolution can be greatly improved Res
282. new research directions that highlight the power savings for ultra low duty cycling and also concludes that the power consumption of SCP de creases with faster radios but that of LPL increases As in energy harvesting appli cations during data forwarding it may not be able to find forwarding nodes as they may be out of current radio range or might not be able to respond because the radio is disabled or battery is recharging Therefore we need to provide queuing of data messages carrying payloads be able to build new neighborhood lists proactively ac cording to the rechargeable time window and be able to sequence the data values and timestamp from individual nodes so that the values can be aggregated as they arrive in the same sequence from source to sink A service based architecture is used that makes writing applications transparent to the data delivery mechanism in high traffic or adaptive burst wireless traffic conditions 12 1 2 RTOS Abstraction Layer Whenever your application handles a variety of activities or manages multiple devices a real time operating system RTOS can help you simplify the code by separating different tasks An RTOS is thus an important tool that allows you to divide and conquer At its simplest level an RTOS is just a context switcher plus a mechanism for han dling some intertask synchronization see RTOS module configuration in Table 12 1 The context switcher allocates the CPU to various tasks according t
283. ng the privacy of medical data research has been conducted specically for this type of application CodeBlue is a project at Harvard University to explore wireless sensor network technology in arange of medical applications such as disaster response prehospitalization and in hospital emergency care and stroke patient rehabilitation In this project a software infrastructure was developed to provide routing discovery and security for wireless medical sensors PCs PDAs and other devices so that patients can be monitored and treated in a wide range of medical settings 2 1 4 Vehicle Management Inexpensive wireless sensor networks enable many applications in vehicle manage ment such as traffic control vehicle tracking and detection monitoring car theft and parking management The current vehicle traffic monitoring systems e g see Fig 2 2 use some buried sensors cameras and an associated communication network which are expensive and generally limited to a few critical points In the future cheap sensors with networking capability can be attached on in vehicles and deployed at every road intersection Thus traffic information such as the speed and density of traffic and the location of traffic jams may not only be obtained by the traffic control center but also exchanged among vehicles passing each other Further sensors in stalled in a vehicle and communicating with those in other vehicles can help reduce the likelihood of veh
284. nherent to the understanding of the technology of sensor networks S S Iyengar Wireless sensor networks have the potential to become significant subsystems of en gineering applications Before relegating important and safety critical tasks to such subsystems it is necessary to understand the dynamic behavior of these subsystems in simulation environments There is an urgent need to develop simulation plat forms that are useful for exploring both the networking issues and the distributed computing aspects of wireless sensor networks Current efforts to simulate wire less sensor networks focus largely on the networking issues These approaches use well known network simulation tools that are dificult to extend to explore distributed computing issues This chapter presents an architecture of a sensor simulator and a sensor node that is used in the simulator This chapter further emphasizes that OMNeT4 is a viable discrete event simulation framework for studying both the networking aspects and the distributed computing aspects of sensor networks We present the architecture of a sensor node that is used in the simulator and the general architecture of the simulator On the basis of our studies with the IEEE 802 11 MAC and directed diffusion in tegrated with GEAR we conclude that our simulator is at least an order of magnitude faster than ns 2 and uses memory more eficiently than ns2 The modular structure of compound modules and the ease of c
285. nloads can typically be handled in a background mode using a best effort protocol The bridging function which transports data between multiple networks is important in contemporary distributed systems such as DSNs that are likely to be integrated into existing engineering systems Bridging refers to tasks performed on interface devices that connect two or more networks The protocol used on the networks may or may not be the same These intelligent devices provide services such as data filtering data fusion alternate routing and broadcasting and serve to partition the system into logical subsets A communication protocol definition such as that in the open systems intercon nection OSD is designed as layers of services from low level physical implementa tion to media access through networking up to the application layer Such layered communication protocols are unlikely to be implemented in resource constrained sensor nodes For this taxonomy we focus only on the media access communication MAC layer since it appears to be the layer where most variations occur Under the MAC protocol implementation attributes we consider two attributes the address ing scheme and the access mechanism The method of addressing messages called the addressing scheme can be source based in which only the producing device s address is used in messages versus using the destination address to route the mes sage Source based schemes can be extended to us
286. nnel that leads to the aircraft The other scanner placed on the loop with the help of diverters dispatches bags for further inspection or diverts them to a channel that leads to the aircraft 2 Leone et al emphasize the importance of relying on computer based models to simulate the impact of integrating explosive detection devices at checkin terminals in airports on the overall operation To better understand the influence of the checked baggage screening CBS process on the overall operation a five stage model is www finebook ir BACKGROUND 279 Desk 1 Desk 2 Desk 2 Level 4 gt Inspection Induction into BHS XRAY XRAY b P Scanner f ji Scanner Scanner Sorting Z Diverter 1 Waiting Aircraft FIGURE 15 2 Baggage checkin proposed as shown in Fig 15 3 This logical view along with other parameters that govern operation e g ticket counter checkin processing time will serve as input in developing a model to study demand and capacity requirements 4 Le et al are concerned mainly with finding the optimal input conditions to ensure optimal throughput performance Their paper presents an empirical study of a mul tiobjective problem within a BHS with a goal of estimating the near optimal input conditions such that no blockage occurs at checkin stations while minimizing the baggage travel time and maximizing the throughput performance measures They provide
287. nodes can be divided into three cat egories augmented general purpose computers dedicated embedded sensor nodes and system on chip nodes A sensor node of the first type normally has a higher volume than do the other two types and relatively higher processing and memory capabilities Off the shelf operating systems such as Linux and Win CE are run in real time and standard wireless communication protocols such as IEEE 802 11 or Bluetooth are used in the node A wide range of sensors from simple microphones to sophisticated videocameras can be accommodated on the node Examples of this type of nodes include Sensoria WINS sGate nodes and various personal digital assistant PDA devices Dedicated embedded sensor nodes have a low compact volume as one of the design objectives They have limited processing and memory capacities Thus special operating systems such as TinyOS and companion programming languages have been developed Sensor nodes of this type include the Berkeley mote family and SunSPOT Because of the low cost and low volume features the future widespread use of such nodes is expected A system on chip node appears to be the direction of future sensor nodes where extremely low power and small size are achieved Such types of nodes will enable new applications to wireless sensor networks It is envisaged that for example the tiny nodes can be mixed in the paint material and painted on the surface of a bridge to monitor the safety of t
288. npredictability of sensor nodes and links in sensor networks leads to sensor data uncertainty Sensor data might not be delivered at reliable rates the data may be incorrect as a result of sensor node damage incomplete because of packet loss or inaccurate owing to environmental interference Techniques have been developed to derive proper information from the data obtained with uncertainty Heterogeneous Sources Data from heterogeneous information sources pose chal lenges to data processing techniques in wireless sensor networks In practice different sensors will be deployed to obtain information from different sources so different data formats and semantics should be considered in data processing Data from static data sources may be used along with sensor data for answering queries PROBLEMS 2 1 List and discuss some WSN applications 2 2 Briefly discuss the three types of sensor nodes 2 3 Explain how you would set up a real wireless sensor network to monitor the temperatures in four corners of a room List the equipment sensor boards base stations interfaces etc that you would use and detail the steps you would follow to set up the sensor network 2 4 Wirite a one page summary of the article by Szewczyk et al 4 2 5 Prepare a one page summary of the article by Lorincz et al 5 www finebook ir REFERENCES 25 2 6 Write a two paragraph 500 word summary of the article by Essa 6 2 7 Discuss the characteristics of
289. ns preemptively in the sense that it can be preempted before completion All async elements can only interact with other async components For example an asynchronous command can only call other asynchronous commands When a need to signal an event arises signaling is synchronous in nature the asynchronous command can post a task asynchronous in nature which then signals the event By default all code written in nesC is synchronous but the addition of the keyword async denotes it as being asynchronous Asynchronous operations usually have short execution cycles Some examples of asynchronous elements in TinyOS are interrupt handlers In the following sections we discuss the main drawback of having asynchronous elements and ways to mitigate it 6 2 3 Preemption Problems One issue with async functions is that when an async function is preempted by another function modifying the same shared variables it can affect the underlying computation producing erroneous results Consider the following example www finebook ir 106 PROGRAMMING IN NesC int x 1 async command int add x x 1 interrupt at this line 1 if x 2 0 return TRUE else return FALSE When this function runs without interruption for the first time it returns TRUE Suppose that it is preempted at line 1 by itself then the new call of the function will increment the value of x again making it 3 and the new call will return FALSE After this
290. nsor Technology 3 1 Sensor Level 27 3 2 ServerLevel 33 www finebook ir 12 5 xiii xvii xix xxi XXV 15 27 vii viii II III 7 CONTENTS 3 3 Client Level 36 3 4 Programming Tools 36 Problems 37 References 38 BACKGROUND Data Structures for Sensor Computing 41 41 Introduction to Sensor Computing 43 4 2 Communication Capabilities 46 4 3 General Structure of Programming 48 4 4 Details on Embedded Data Structures 51 4 5 LinkedList 53 4 6 Importance of Graph Concepts in Sensor Programming 57 4 7 Graph and Trees 61 4 8 Trees 66 4 9 Graph Traversal 75 4 10 Connectivity 76 4 11 Planar Graphs 81 4 12 Coloring and Independence 83 4 13 Clique Covering 84 4 14 Intersection Graph 85 4 15 Defining Data Structure of Spanning Tree Protocols 86 Problems 90 References 91 Tiny Operating System TinyOS 92 5 1 Components of TinyOS 93 52 An Introduction to NesC 93 5 3 Event Driven Programming 96 Problems 97 References 97 Programming in NesC 99 6 1 NesC Programming 99 6 2 A Simple Program 99 Problems 108 References 109 SENSOR NETWORK IMPLEMENTATION Sensor Programming 113 7 Programming Challenges in Wireless Sensor Networks 113 www finebook ir 8 IV 10 CONTENTS ix 7 2 Sensing the World 119 7 3 Applications Using the Interface SplitControl 122 Problems 129 References 130 Algorithms for Wireless Sensor Networks 131 8 1 Structural Characteristics
291. nts new TimerMilliC algorithm2C Boot MainC Boot algorithm2C Random RandomC algorithm2C Leds LedsC algorithm2C AMPacket AMSenderC algorithm2C Packet AMSenderC algorithm2C AMSend AMSenderC algorithm2C AMA ActiveMessageAddressC algorithm2C PacketAcknowledgements AMSenderC www finebook ir 174 TECHNIQUES FOR PROTOCOL PROGRAMMING algorithm2C Receive AMReceiverC algorithm2C RadioControl ActiveMessageC algorithm2C Timer0 TimerMilliC J void generateAddress call Leds set 0 my address 31071334523 c a 1 l Random randl6 void broadcastAddress message t msg am addr t temp temp am_addr_t call AMSend getPayload amp msg temp call AMA amAddress call AMSend send 0 x ffff amp msg sizeof amaddrt In addition to the commands and events from the interfaces we also use two more local functions for generating random addresses and broadcasting the addresses to neighboring nodes event void Boot booted call RadioCont rol start i event void RadioCont rol startDone error_t err puts the address in my_address generateAddress The initial address broadcast call TimerO0 startOneShot 6000 Check once if any one else has my address If not at this point we can set the generated address call AMA setAddress my group my address The variable my address is a global variable in the implementation section of the component
292. nvironments for large number of low impact sensors Similarly motor vehicles which already use large number of microprocessors in their existing systems will be able to easily evolve to include sensors and wireless communication devices as standard hardware The effect of this will be a potentially information dense environment with over whelming amounts of data available from anywhere inhabited by humans In the long term the greatest cost and the challenge involved in building sensor networks will rest not in the hardware but in the software In order to deal with large numbers of nodes measuring large amounts of data whole new paradigms in data aggregation and network architecture will be needed Networks on this scale will resemble a living system more than a series of traditional computer networks and our approach to them must vary accordingly The deployment of sensors often necessitates an understanding of certain requirements such as sensor coverage redundancy and network architecture providing the essence of sensing physical space Applications such as multiple target tracking environmental monitoring health monitoring energy usage monitoring or other general security monitoring require varying degrees of sensor coverage Certain networking abstractions must be created to ensure easy management of sensors irrespective of the coverage type dense Vs less dense networks In the following sections we discuss some of the common abstracti
293. nvolve the dynamics of both the application and the sensor network Such simulation studies must explore the effects of scale density node level architecture energy efficiency communication architecture failure modes at node and communication media levels system architecture algorithms protocols and configuration among other issues Unlike traditional computer systems it is not sufficient to simulate the behavior of the sensor network in isolation because of the tight and ubiquitous coupling between the sensor network and its application 14 2 WHY A NEW SIMULATOR Inarecent report 1 the following paragraph summarizes the need for a new simulator ns2 perhaps the most widely used network simulator has been extended to include some basic facilities to simulate Sensor Networks However one of the problems of ns2 is its object oriented design that introduces much unnecessary interdependency between modules Such interdependency sometimes makes the addition of new protocol models extremely difficult only mastered by those who have intimate familiarity with the simulator Being difficult to extend is not a major problem for simulators targeted at traditional networks for there the set of popular protocols is relatively small For www finebook ir CURRENTLY AVAILABLE SIMULATORS 255 example Ethernet is widely used for wired LAN IEEE 802 11 for wireless LAN TCP for reliable transmission over unreliable media For sensor networks howe
294. o a scheduling algorithm The tasks can therefore advance at different paces depending on how many CPU cycles they obtain An RTOS also provides mechanisms that allow the tasks to synchronize their activities Examples of such mechanisms include various types of semaphores mailboxes and message queues Although the basic services provided by RTOSs are very much the same they are accessible through significantly different APIs This poses a serious problem for many companies that want to deploy shared application code on different operating systems or that don t want to lock their strategic application into a particular operating system An active object based framework such as the INSPIRE offers a more elegant solution 12 1 3 Minimal Application How software is implemented in sensor networks is essential to understanding sensor networks A network architecture and protocols are essential foundations for building software applications Active objects in microframework are encapsulated tasks each embedding a state machine and an event queue that communicate with one another TABLE 12 1 Real Time OS Module Configuration Coresident OS Tasks Priority Number of Control Points Kernel Control Highest 6 digital I O ports Kernel Scan Baud rate sensitive Read only www finebook ir 218 INSPIRE asynchronously by sending and receiving events Within an active object events are processed sequentially in a run to completion RTC fashion while
295. ode on one subnet many nodes on the other sub node may become unreachable While address management is a difficult problem we will discuss a few simple techniques and their programming issues 9 9 DISTRIBUTED ASSIGNMENT OF NETWORKWIDE ADDRESSES This is a simple technique for assigning addresses to each node but the technique is expensive involving too many send operations In this technique each node generates a random address sets it as its own address and broadcasts to its neighbors to check whether it clashes with any of its neighbours If anyone reports address collision the node again tries with another random address and broadcasts to its neighbors again This continues until no clashes occur module algorithm2C uses interface Boot uses interface Random uses interface Leds uses interface AMPacket uses interface Packet uses interface AMSend uses interface ActiveMessageAddress as AMA uses interface PacketAcknowledgements www finebook ir DISTRIBUTED ASSIGNMENT OF NETWORKWIDE ADDRESSES 171 uses interface Receive uses interface SplitControl as RadioCont rol uses interface Timer lt TMilli gt as Timer0 implementation void generateAddress void broadcastAddress am group t my group 1 am addr t my address bool conflict event void Boot booted call RadioControl start S EE K k k k k k k k k k k k k k k k RR RR RRR RRR RR RR K K KOR R K K ke RK K R RRR R R Comput
296. ointer One 1 Two Three 8 www finebook ir 56 DATA STRUCTURES FOR SENSOR COMPUTING TABLE 4 6 Linked List Data Structure Name Pointer One 1 Two 6 Three 9 the linear list T shown in Fig 4 9a and the equivalent representation in linked lists shown in Fig 4 9b The binary search functions see Table 4 7 can be obtained by computing midpoints of the list The following discussion describes the binary search process The midpoint between array elements 1 and 5 in the linear list given above is array element 3 But in the linked list there is no way to find the node midway between two other nodes so a binary search is impossible The most obvious reason to use a linked list framework data structure such as the one described above is the ability to insert or delete dynamic events during the event detection process of a sensor network This framework also frees application programmers from dealing with buffer related communication issues and allows them to concentrate on robust methodologies for complex applications 4 5 2 Circular Lists In circular lists the pointer cell of the last node is not null It points the first node This is shown in Fig 4 10a 4 5 3 Doubly Linked List The basic difficulty of a linked list or circular list is that we may move in only one direction To move in both directions we introduce the concept of a doubly linked list The nodes in a doubly linked list have two links LLINK for the left link a
297. oles employed in a sensor network communication plays a critical role in determining life span routing speed and ultimately the nature of data that can be communicated in wireless sensor networks It is for this main reason that we discuss some of the available protocols and features supported by most MAC implementations 4 1 The S MAC is a medium access control protocol for wireless sensor networks It has been optimized for wireless sensor networks offering some advantages of 802 11 based MAC implementations such as reduced energy consumption and sup port for self configuration In its development three sources of energy waste were identified and improved resulting in energy savings of 2 6 times over those of tra ditional 802 11 like MACs Two of the three areas of improvement are discussed below Collision When two or more nodes try to transmit packets at the same time collisions may occur resulting in packet corruption Over a period of time these collisions represent a significant source of energy drain due to the retransmis sions necessary after packet loss Overhearing In the case of over hearing nodes listening to a channel pick up packets not intended for themselves thus consuming more energy With the S MAC implementation neighboring nodes form virtual clusters and autosynchronize on sleeping schedules thus reducing occurences of overhearing Fundamentals of Sensor Network Programming Applications and Technology B
298. om go permission Limit of Liability Disclaimer of Warranty While the publisher and author have used their best efforts in preparing this book they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose No warranty may be created or extended by sales representatives or written sales materials The advice and strategies contained herein may not be suitable for your situation You should consult with a professional where appropriate Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages including but not limited to special incidental consequential or other damages For general information on our other products and services or for technical support please contact our Customer Care Department within the United States at 800 762 2974 outside the United States at 317 572 3993 or fax 317 572 4002 Wiley also publishes its books in a variety of electronic formats Some content that appears in print may not be available in electronic formats For more information about Wiley products visit our web site at www wiley com Library of Congress Cataloging in Publication Data is available ISBN 978 0470 87614 5 Printed in Singapore 10 9 8 765 43 2 1 www finebook ir This book is dedicated to Professor Donald E Knuth Professor Emeri
299. on Operating Syst Rev 28 4 56 63 Oct 1994 Srinivas M B V Iyer G Rama Murthy and B Hochet C error simulator for development for sensor and location aware sensing applications Proc 3rd Int Conf Sensing Technology Taichung Taiwan 2002 pp 799 804 Stankovic J A Perrig and D Wagner Security in wireless sensor networks Commun ACM 47 53 57 2004 Tannenbaum A S Computer Networks Prentice Hall 2002 Vargas A Omnet Discrete Event Simulation System version 2 3 2003 Vieira M A M D C da Silva Jr C N Coelho Jr and J M da Mata Survey on wireless sensor network devices Proc IEEE Conf Emerging Technologies and Factory Automation ETFAO3 2003 p 1 Wallace C P Jensen and N Soparkar Supervisory control of workflow scheduling Proc Int Workshop on Advanced Transaction Models and Architectures 1996 Wood A and J A Stankovic Denial of service in sensor networks IEEE Comput 35 10 54 62 Oct 2002 Xavier C and S S Iyengar Introduction to Parallel Algorithms Wiley 1998 Ye W F Silva and J Heidemann Ultra low duty cycle MAC with scheduled channel polling SenSys 06 Proc 4th Int Conf Embedded Networked Sensor Systems ACM NewYork 2006 pp 321 334 Yu Y R Govindan and D Estrin Geographical and Energy Aware Routing A Recursive Data Dissemination Protocol for Wireless Sensor Networks Technical Report Aug 2001 Zeng X R Bagrodia and M Ge
300. onfiguring simulation scenarios via an initializa tion file offers us a tremendous amount of flexibility to model and study the dynamic behaviors of both the sensor network and the application environment in which such networks are expected to operate Discrete event simulation is a trusted platform for modeling and simulating a va riety of systems We discuss results obtained from a simulator for wireless sensor networks that is based on the discrete event simulation framework called OMNeT The authors would like to acknowledge the contributions of the former graduate students C Mallanda A Suri and V Kunchakarra and of Drs R Kannan A Durresi and S Sastry This work was supported in part by NSF ITR under IIS 0312632 and IIS 0329738 and DARPA AFRL grant F30602 02 1 0198 Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 253 www finebook ir 254 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION This framework develops simulations for the 802 11 MAC and the well known sen sor network protocol called directed diffusion The performance of the simulator is deomnstrated by comparing its performance to that of the well known simulator ns 2 Results indicate that the OMNeT simulator executes at least an order of magni tude faster than ns 2 and makes more efficient use of the av
301. ons used to facilitate the deployment and management of sensors 10 1 SENSOR NETWORK ABSTRACTION Sensor networks should have the following properties Self configuration formation of networks with no human intervention Self healing automatic deletion addition of nodes without resetting the entire network Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 197 www finebook ir 198 SENSOR DEPLOYMENT ABSTRACTION FIGURE 10 1 Data aggregation flow diagram Dynamic routing adapting routing schemes on the fly on the basis of network conditions such as link quality hop count and gradient Multihop communication improving the scalability of the network by sending messages on a peer to peer basis to a base station In order to support these functionalities at the large network level programming design patterns and abstractions are essential We will now present a few of them 10 2 DATA AGGREGATION In network aggregation of data see Fig 10 1 is one of the fundamental data processing techniques commonly used in sensor networks This is because significant energy savings can be accrued over time as nodes collectively collaborate to forward data toward a data sink rather than having each node perform this task single handedly In TinyOS some networking components having
302. ook ir 230 INSPIRE 12 1 8 Application with Sensing This category of applications pools the sensors on a periodic basis which allows average values to be accumulated over time With this facility the application can control the service usage which further allows the service to adapt to predetermined data sensing requests and also to redundant traffic void main void ifndef LRWPAN_COORDINATOR UINT8 count BOOL aps ack UINT8 failures UINT32 my timer ftendif Hallnit evblnit will have to be called by the user hallnit evblnit apllnit init the stack conPrintConfig ENABLE GLOBAL INTERRUPT enable interrupts EVB LED1 OFF EVB LED2 OFF debug level 10 ifdef LRWPAN_RFD rfdState RFD STATE JOIN_NETWORK while 1 apsFSM switch rfdState case RFD STATE JOIN NETWORK EVB LED1 OFF not connected to a network aplJoinNetwork rfdState RFD STATE JOIN WAIT break case RFD STATE JOIN WAIT if apsBusy break www finebook ir MOTIVATION AND BACKGROUND 231 if aplGetStatus LRWPAN STATUS SUCCESS conPrintROMString Network POR ENS RE RET NEE Tat Soe tree aT Join succeeded n printJoinInfo rfdState RFD STATE NORMAL ping ent 0 count 0 aps ack FALSE EVB LED1 ON j else conPrintROMString Network sete Lath al ah aed a a galt ar hang Join FAILED _ conPrintROMString Error conPrintUINT8
303. oordinator but since don t know the network topology do it always if RFD do send to coordinator as it resolves bindings aplSendEndDeviceAnnounce 0 www finebook ir MOTIVATION AND BACKGROUND 223 wait for finish while apsBusy apsFSM if aplGetStatus LRWPAN STATUS SUCCESS conPrintROMString End Device Meo Announc succeeded n break else conPrintROMString End Device Mis Goel Cad EHE Ex CIC ER eee Announce FAILED jh M es Waiting then trying again un my_timer halGetMACTimer wait for 2 seconds while halMACTimerNowDelta my_timer lt MSECS TO MACTICKS 2 1000 while 1 endif if defined LRWPAN RFD defined LRWPAN COORDINATOR now send packets while 1 packet test while apsBusy apsFSM wait for finish endif ifdef LRWPAN ROUTER router does nothing just routes conPrintROMString Router TERR RAE CUN Maes Meaney doing its thing Wn www finebook ir 224 INSPIRE while 1 apsFSM endif void main void STACK API Program this initializations et our SADDR to OXFFFF PANID to the default PANID Hallnit evblnit will have to be called by the user hallnit evblnit apllnit init the stack conPrintConfig ENABLE GLOBAL INTERRUPT enable interrupts test number 0 debug level 10 ifdef LRWPAN
304. or components SightC components LedsC SightC Boot MainC SightC SightControl ActiveMessageC SightC Timer Timer0 SightC Read gt Sensor SightC Leds LedsC 7 3 2 PacketSender This is a simple example where we send a packet of data from one mote sensor node to another mote and also illustrate how to use the three intrfaces Packet AMsend and Receive This is the first example where we attempt communication from one node to another Fig 7 4 shows node A transmitting a packet of data to node B In the program we discuss below one node creates a packet consisting of an integer number sends it to a neighboring node and waits for a message from it We load a copy of this application program in each node It uses the LEDs to display the intermediate states of execution indicating the end of subtasks that are achieved successfully as the application continues to run It should by now have become clear to us that we will be needing the following interfaces Boot to boot start application Leds to display the intermediate states of the execution LEDs are the only output devices we have for the human to look at and understand what is happening inside the sensor nodes and Timer to run a clock that can generate clock pulses to fire events We will also need interfaces for communication purposes The interface AMSend can be used for sending and Receive for receiving packets of messages We need one interface that can be used to manipulate
305. or networks can change rapidly and frequently The nodes in a sensor network do not have a global identifier such as an IP Internet Protocol address and the number of sensor nodes in a sensor network may be an order of magnitude greater than that in a typical computer network The memory and the processing capabilities of sensor nodes are limited in comparison to nodes in a computer network These characteristics lead to a programming environment that is unique Thus the programs need to be short and efficient providing capabilities of interfaces and links of components and modules to each other Additionally to save battery power the nodes may need to have aggressive power management capabilities thus the programming environment needs to provide mechanisms such as split phase the nonblocking equivalent of common power saving techniques such as the sleep command 16 GOALS OF THE BOOK The goals of this book are to develop programming methodologies unique to sensor networks and present in an organized fashion techniques for programming of sen sors to enable them to work effectively as a group Thus although the focus is on programming of the individual sensor the goal is to enable the sensor to work within a collaborative environment 17 WHY TinyOS AND NesC TinyOS is an emerging platform that provides a framework for the most common type of sensor application programming Thus we have a tool that can be implemented on small Crossbow sen
306. orks CRC Press 2002 C Xavier and S S Iyengar Introduction to Parallel Algorithms Wiley 1998 Ww N aD www finebook ir 5 Tiny Operating System TinyOS Simple things should be simple complex things should be possible Alan Kay An operating system OS can be defined as the master program existing between the computer hardware and the user Its main purpose is the management and coor dination of various system resources which may include hardware resources such as hard disk memory peripheral device management or nonhardware resources such as managing processor timeslots It provides equal access to these resources through numerous system interfaces Also included in its role as a master program an oper ating system can act as a host on which other programs can be executed providing systemwide services by which these programs can access shared resources Typical services provided by operating systems include networking services file I O data security and provision of user interfaces There are several types of operating system each differing only in its intended environment for use Some common examples of operating systems include real time embedded single user and multiuser operat ing systems In most wireless sensor networks sensors implement a basic operating system containing the minimum functions necessary to execute their tasks This min imalist approach is of great importance because of the highly res
307. orks providing information on how large scale mobile ad hoc wireless networks can be created and managed efficiently 2 Power Systems Sensor networks also rely on information from computer science and electrical and nanosystems engineering in the creation of energy efficient software and hardware components resulting in improved life of sensor networks 3 Data Management Experience in large scale data management and data mining techniques is required in sensor networks since huge heterogenous datastreams are generated from these ubiquitous sensing devices 4 Data Fusion Since most devices have basic sensing capabilities the need to create software systems capable of combining data from multiple sources to create more complex representation of the world is necessary hence the need for data fusion Fusion systems draw on advances in artificial intelligence statistical analysis and distributed systems 1 2 3 Technology Driven Methods A few examples of technology driven methods in sensor networks follow 1 Flooding such as broadcast of packets in a synchronized network from source to destination until the path is formed to find the topology 2 Clustering including K means clustering to find K centers and form a cluster to minimize the distance between nodes in a dense region and efficiently form a topology 3 Short path algorithms for data aggregation such as data aggregation trees to form wireless spanners to efficiently coll
308. orm clusters and uses efficient clusterhead selection and the other at the MAC layer that uses multihop routing to forward data at the lower layers by using best effort quality of service QoS As there are many different radios available the need to make them work within a given framework is an integral part of the basic modular design As the requirement TABLE 12 3 STACK APIs Protocol Stack Fully Functional Routing Stack Monitoring Only Layers Stack Stack Application Form Network Do Routing Join Network Create Zones Elect Measure API ClusterHeads Datalink CRF CRF CRF Physical IEEE Standard IEEE Standard IEEE Standard OS scheduler Create HostComm Compress Create SenseTask Microframework Create SenseTask 2 www finebook ir MOTIVATION AND BACKGROUND 221 is to develop a servicelike application that can control the MAC we design all the core services needed to make it protocol efficient Any built in features refer to Optimized for power aware OS This is the case as the ultimate goal of the overall system in which any application is deployed is to have a longer lifetime void main void Coordinator Program this initialization set our SADDR to OxFFFF PANID to the default PANID Hallnit evblnit will have to be called by the user hallnit evblnit apllnit init the stack conPrintConfig ENABLE GLOBAL INTERRUPT enable interrupts te
309. ource constrained nature of these sensing devices operating system code and application code must re side in devices having much less than one megabyte 1 MB of memory Sensor based operating systems typically provide two basic services a hardware abstraction layer for accessing the various sensing devices and a network interface for communication There are several sensor based OSes in use today some of these are TinyOS Contiki Mantis OS BTnut and Nano RK In the following sections we discuss TinyOS and its associated programming language nesC in detail 1 TinyOS as its name implies can be described as a miniature framework designed for embedded systems that require very aggressive resource management due to the highly constrained nature of their resources such as power and available memory 3 It implements the hardware abstraction layer and scheduler of a conventional operating system allowing generic programs that may have no knowledge of the intricate details of the operations supported by the underlying hardware components such as sensors to use well defined interfaces to interact with these components Its C language like framework provides an interface to core system components Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 92 www finebook ir AN INTRODUCTION TO NesC 93
310. ove only the topmost entity The simple procedure for adding a new entity to a stack is shown below Procedure STACK ADD S 1 n Top item BEGIN IF top n then call STACK FULL ELSE top lt top l S top item Endif Procedure STACK DEL S 1 n top item BEGIN IF top 0 then call STACK EMPTY ELSE item S top top top 1 Endif www finebook ir LINKED LIST 53 Queue Rear e Queue Front FIGURE 4 6 A queue In this procedure it is assumed that the maximum capacity of the stack is n If the stack is already full we cannot insert a new item in the stack So if top n we call a procedure stack full which we assume will eliminate the need for further processing In a similar way we can write a procedure for deleting one item from the stack If top 0 we assume that the stack is already empty In this case we cannot delete anything from the stack The stack data structure has numerous applications in algorithms both sequential and parallel The prefix postfix infix representation of arithmetic expressions recur sive procedures scheduling and polynomial evaluation are some important areas of application of the stack data structure 4 4 4 Queue The queue see Fig 4 6 is an array in which the additions take place through one end called rear and deletions take place through the other end called front In order to make the queue a
311. ow causing nodes in a continually expanding www finebook ir DENIAL OF SERVICE ATTACKS IN MULTIPLE LAYERS 301 Black Hole FIGURE 16 3 Packets being routedtoa black hole radius to route their data toward the black hole see Fig 16 3 This attack can be detected by looking for suspicious routing cost claims and if detected is fairly easy to defend against However if not detected it can be extremely disruptive Having nodes monitor their neighbors provides a good defense against this type of attack although with overhead Occasional network probing can also be used to locate blackout areas and distributed probing schemes are possible It is important that a probe be indistinguishable from ordinary traffic or a malicious node could appear benign during probing Flooding is when many packets are sent to a node in an attempt to overload it One example of this is the SYN flood where an attacker sends a large number of requests for connections binding up all the target s resources on pending transac tions Similar types of attack are possible on WSNs where an adversary can waste a node s resources by sending many connection requests Limiting the number of incoming connections will save resources but service of legitimate connection re quests will still be slow at best A better option is to make it computationally costly for a node to make a connection establishment request Client puzzles are simple mathematical puzzles
312. pation and loading of traffic at each node for different algorithms This treats the large network as a testbed with standard metrics used by life time and standard AA cells The testbed simulator allows us to compare the performance characteristics of different hardware MAC protocols and their energy requirements 13 1 1 Performance Metrics We can define lifetime of sensor networks as Coat x V x 60 x 60 Node lifetime et D P x 100 13 1 Tx Rx idle Lis where C power consumption E energy V voltage Rx Tx receving transmitting and Lispassive Idle power consumption is defined as the time elapsed between request and receipt of information over the wireless channel This is increasing at an alarming rate due Portions of this chapter were contributed from various sources by V Iyer 1 4 Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 239 www finebook ir 240 PERFORMANCE ANALYSIS OF POWER AWARE ALGORITHMS Energy Wastage of 100 nodes with 2 packets sec sec E B MAC 78 75 Radio 0 4 0 09 0 03 IDLE Rx Tx Sen FIGURE 13 1 Energy wastage due to radio using Eq 13 1 to exponential growth of idling Figure 13 1 illustrates the simple WSN application that aggregates data with 100 nodes that must run the radio of all the nodes all the ti
313. pect to grid coordinates Then we identified a fixed number of query generating nodes and distributed these nodes randomly in the sensor field Next we determined a target region and specified the boundary of the region in terms of the grid coordinate and the number of sensor nodes in the region We then executed the simulation for a specified duration and observed the ratio of the number of packets generated in the region to the number of packets received by the query generating nodes The IEEE 802 11 MAC was considered at the MAC layer with a simple pass through the physical layer for these simulations The results for 5 2000 nodes shown in Fig 14 6 for SensorSimulator and ns 2 indicates that for a similar topology and simulation environment the delivery ratio is comparable with ns2 www finebook ir 266 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION 101 100 99 98 SensorSimulator B ns2 Delivery Ratio N 5 10 50 100 200 Number of nodes FIGURE 14 6 Comparison of delivery ratio We also executed the simulations to verify the directed diffusion implementation on our sensor simulator by observing the changes in delivery ratio of data packets by region by varying the number of queries as shown in Fig 14 7 Number of nodes 500 Simulation time 300 seconds Network size 500500 Meter square area Number of nodes in region at least 10 non faulty nodes per cluster 14 6 2 Directed D
314. performance is to have receiver centric duty cycling for the majority of the nodes thereby minimizing the number of active required nodes in the connected network That is precisely how the INSPIRE runtime framework process works as shown in the Eq 13 2 with the duty cycle as shown in Fig 13 2 Coat x V x 60 x 60 Node lifetime st 5 gt x duty cycle 13 2 Tx Rx idle Lis www finebook ir INTRODUCTION 241 EN Deu Rx with Sleep Idle 200r 180 160r 140r 120r 100r 80r 40 20 0 H m H R A R Li Routing rounds completed during lifetime of the Radio power Amps consumed after 20 000 simulation seconds FIGURE 13 2 Unbalanced energy power resource allocations due to radio using Eq 13 2 The INSPIRE process utilizes several innovative technologies to extend the life time of WSN applications One advantage of these techniques that specifically ad dresses unbalanced resource allocation for a proactive application is that it is very sensitive to Tx energy cost As noted in Fig 13 1 the Tx peaks during the application lifetime This runtime environment FARMS uses a renewable energy model that attempts to average the resource allocation over the entire lifetime for ultra low duty cycle applications so we rewrite Eq 13 2 to use ultralow duty cycling as shown in Eq 13 3 and add a rechargeable cycle to renew ambient energy as given in Eq 13 4 thus making new applications sign
315. pletion of an operation in the split phase strategy is signaled by invoking a corresponding event Initially it might seem harmless to signal an event from a different component at the end of the code that implements the operation this is what we have done in PrintMessageC nc above Serious problems can sometimes arise with this approach For example if the code for a command C makes a direct call to its corresponding event and the event in turn makes a call to the command C then this might cause a potentially long loop causing the system stack to overflow We can avoid this by using a task which is a deferred procedure call 1 The command C instead of signaling its corresponding event e immediately can create a task that signals the invocation of the event e and then post it on the system queue The task will eventually be executed thus potentially avoiding the creation of a long loop Print MEssageC below shows how to implement sendDone as a task module PrintMessageC provides interface PrintMessage implementation command error_t PrintMessage send char msg 20 uint8 t len error_t err if strlen msg lt 20 amp amp len lt 20 dbg std Message was s n msg err SUCCESS post sendDoneTask msg err return err else err FAIL post sendDoneTask msg err return err end commad end implementation task void sendDoneTask char msg 20 err 1 Signal PrintMessage sendDone msg er
316. pp 104 111 S Sastry Smart space for automation Assembly Autom 24 2 201 209 2004 S Sastry S S Iyengar and N Balakrishnan Sensor technologies for future automation systems Sensor Lett 2 1 9 17 2004 A Sobieh and J C Hou A similation Framework for Sensor Networks in j sim Technical Report UIUCDCS R2003 2386 Dept Computer Science Univ Illinois Urbana Champaign Nov 2003 A Tannenbaum Computer Networks Prentice Hall 2002 A Vargas Omnet4 4 Discrete Event Simulation System version 2 3 2003 Y Yu R Govindan and D Estrin Geographial and Energy Aware Routing A Recursive Data Dissemination Protocol for Wireless Sensor Networks Technical Report Aug 2001 X Zeng R Bagrodia and M Gerla Glomosim A library for parallel simulation of large scale wireless networks Proc Workshop on Parallel and Distributed Simulation 1998 pp 154 161 www finebook ir 15 MATLAB Simulation of Airport Baggage Handling System If we want users to like our software we should design it to behave like a likeable person Alan Cooper 15 1 INTRODUCTION This section discusses a simple implementation of an intelligent baggage handling system in MATLAB Airport baggage handling systems BHSs fall under the broad umbrella of material handling systems that automate the process of moving mate rials Current BHS controllers are based on standard industrial techniques namely Programmable Logic Controllers PLCs
317. pplication with Sensing This category of applications pools the sensors on a periodic basis which then allows average values to be accumulated over time As a result of this feature the application can control the service usage which further allows the service to adapt to a predetermined data sensing request and adapt to redundant traffic 13 2 6 Ultra Low Duty Cycling Using FARMS When energy resources are abundant owing to renewable energy resources the run ning harvesting application is sensitive to the recharge rate As this allows it to transmit data to a forwarding node the communication need not be coordinated however this feature enables the application to ensure that the channel is available For reliable delivery there should be a sufficient number of receivers active so that no forwarded packets are dropped This needs to be scheduled as receiving and idling can drain all the nodes in a given area if they are not actively scheduled This development framework allows us to implement a very optimized version of routing techniques with specific data sensing needs and allows the framework to be transparent to the developer For details on the performance comparison of the power consumption for each application category see Figs 13 12 13 14 13 3 APPROACHES TO WSN PROGRAMMABILITY 13 3 4 GlomoSIM GlomoSIM a commercially available networks imulator combines the simulation kernel which can be run on a dual core PC GlomoSIM scal
318. processing with other nodes The technique employed is the divide and conquer technology in order to provide a robust network reliability for hazardous applications In the following sections we provide a structured approach in terms of describing the new computational system INSPIRE in detail 12 1 1 INSPIRE An Introduction Our system INSPIRE innovation in sensor programming implementation for real time environments adopts an active sensor approach to allow any distributed algo rithm to be executed in the network Energy is a critical resource in sensor networks MAC protocols such as S MAC and TMAC 6 coordinate sleep schedules to reduce energy consumption More recently low power listening LPL approaches such as WiseMAC and B MAC exploit very brief polling of channel activity combined with long preambles before each transmission saving energy particularly during low net work utilization Synchronization cost either explicitly in scheduling or implicitly in long preambles limits all these protocols to duty cycles of 1 2 We demon strate ultralow duty cycles of 0 1 with the use of a mathematical model 5 for lifetime and also support it with simulation results 4 that highlight the power sav ings due to overhearing and node density radio interference when idling We use the www finebook ir MOTIVATION AND BACKGROUND 217 extended work of a new MAC protocol called scheduled channel polling SCP 8 This work prompts three
319. r PrintMessageC Implementation A task when posted goes into the queue and waits in the queue until it is dequeued and serviced by the system Long running tasks www finebook ir A SIMPLE PROGRAM 105 can be broken down into chunks and reposted on the queue Tasks are defined to be non preemptive with respect to one another In other words if a task is running then any other tasks ready to run will have to wait until the currently running task is completed So if there were any shared variables their values will be protected from any other tasks changing it For this reason it is often advisable that tasks be reasonably small so that they can run to completion quickly in order to allow the waiting tasks to start running 6 2 2 Asynchronous Commands and Events The commands in an interface can be classified into two types Asynchronous Synchronous The async keyword preceding a command name denotes a command that may run in asynchronous context By default all command declarations in TinyOS are synchronous 1 interface PrintMessage multi lt val_t gt async command error t sendi char msg 20 val t gt len async command error t send2 char msg 20 lt val t gt len async event void sendl Done char msg 20 error t err async event void send2 Done char msg 20 error t err This code redefines our earlier interface PrintMessage multi making each command and event asynchronous An async command or event ru
320. r The implementation is described with a block diagram in Fig 14 4 The implementation used a simple pass www finebook ir 262 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION m mr Oo mHOH Z77 UZzxEOOD0 PhysicalSimple Wireless Channel Module Query path from Network Layer to MAC Query path from MAC to Network Layer FIGURE 14 4 Implementation scenario through the physical layer a simple wireless channel module with the application layer generating query packets and forwarding to the network layer 14 5 1 Directed Diffusion with GEAR We have implemented directed diffusion 12 along with geographic routing The application layer generates interests that specify the region the kind of data required and rate of delivery of data Nodes that initiate the interest are called subscribers On receiving the interest message the network layer broadcasts beacon messages in the network The immediate neighbors of the node on receiving beacon messages reply www finebook ir IMPLEMENTATION DETAILS 263 Query attribute Query rate of data Query duration FIGURE 14 5 Structure of a query back with a beacon reply type message that contains their geographic location and the energy left in them On receiving the beacon reply messages the neighbor table of the node that sent the beacon is updated The node waits for a fixed duration of time to receive the beacon reply from all the neighbors
321. ration Questions of where programs and documents re side in the distributed system arise as do issues in version control and concurrent access Finally the degree of transparency in synchronization that is provided by the languages and environment is a key to simplifying distributed programming The language chosen in a DSN to implement the algorithm affects the services and tools that must provide support e g operating system compilers partitioning per formance estimation The IEC 1131 programming standards for digital controllers and the more recent IEC 61499 extensions that define an event driven execution model are interesting considerations for programming DSNs Ladder logic is rel atively simple to learn is easy to use and provides a low level ability to react to process changes Sequential function charts Petri nets and finite state machines FSMs are examples of state based languages An FSM model is intuitively simple but the size of the model grows rapidly as the size of the control system increases Hierarchical representation methods such as hierarchical FSMs have been used to cope with the large size of state based models While such hierarchical methods were well suited for hardware design their use in software design is still an ongo ing research issue Function blocks are designed as a replacement for ladder logic programming in an industrial environment They provide a graphical software IC integrated circuit style
322. ration packet NVA lt currentTime duration sleepUntil NVA print Woken up after sleeping end Node D begin while true packet listen if packet equals RTS break endIf endLoop B is going to send data to C Must send D to sleep by setting NVA Variable duration extract Duration packet NVA currentTime duration sleepUntil NVA print Woken up after sleeping end Broadcast RTS To keep the illustration simple we have provided pseudocode for only one role at each node The nesC pseudocode is given below www finebook ir PROBLEMS 153 Theme Simulation of RTS CTS Handshake In this chapter we have illustrated several protocols and shown how they can be coded in the nesC language We observed that coding the protocols themselves was fairly straightforward Node B booted send RTS to C receive if CTS is received transmit data if ACK is received do nothing Node C booted NIL receive if RTS received send CTS if data received store data if end of data received send ACK Node A booted NIL receive if RTS received then extract duration d1 from the packet lowPowerSleep for duration d1 that is set NAV busy Node D booted NIL receive if CTS is received then extract duration d2 from the packet lowPowerSleep for duration d2 that is set NAV busy oo PROBLEMS
323. ratureC Read We now provide the code for the module SighC module SightC uses interface Boot interface SplitControl as SightControl interface Timer lt TMilli gt interface Read lt uint16_t gt interface Leds implementation void errorReporter int error uint16_t sensor value 0 event void Boot booted errorReporter 1 if call SightControl start SUCCESS errorReporter 1 i event void SightControl startDone error t err www finebook ir 124 SENSOR PROGRAMMING if err SUCCESS errorReporter 0 call Timer startPeriodic TIMER FREQUENCY else i errorReporter 1 event void SightControl stopDone error t err event void Timer fired call Leds led2Toggle if call Read read SUCCESS errorReporter 1 event void Read readDone errort result uint16_t val if result SUCCESS sensor value val call Leds led1lToggle else errorReporter 1 void errorReporter int error if error 1 call Leds ledOOn else call Leds ledOOff Finally the following code for the configuration shows how wiring is done configuration SightAppc implementation www finebook ir APPLICATIONS USING THE INTERFACE SplitControl 125 components MainC components ActiveMessageC components new TimerMilliC as Timer0 components new TemperatureC as Sens
324. re lifetime in real time clocks sensor errors and radio profiling during radio idle receiving and transmissions As the practicality of the sensor network is based on limited power resources we extend this power model by using renewable energy resources and its effects on the performance levels of different algorithms 12 1 MOTIVATION AND BACKGROUND A WASN is an ad hoc network of resource limited static wireless sensor nodes that are used to monitor dynamic physical processes Typically a user queries the Portions of this chapter were contributed from various sources by V Iyer 4 5 Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 215 www finebook ir 216 INSPIRE network the query triggers some reaction from the network and as a result the user receives the information needed The reaction to the query can vary from a simple return of sensor value to a complex unfolding of a distributed algorithm among some of all of the sensor nodes such as a collaborative signal processing algorithm a distributed estimation algorithm or an actuator control algorithm Furthermore there are multiple users who are transiently connected to the network each having different needs in requested information These systems are quite different from traditional networks 1 they have severe energ
325. rface table command build event buildDone err module floodC uses interfaces initialize read flood table booted FLOODING Pseudo code Source role read data and flood the network sink role reveive data process intermediate node rol www finebook ir 184 TECHNIQUES FOR PROTOCOL PROGRAMMING if received first time flood the network pseudo nesC Components Assume Let T be the neighborhood table booted self initialize role firstTime true if self source then read x readDone d data from read operation flood the network receive if self sink then extract packet and process if self intermediate amp firstTime then extract data flood the network firstTime false In this example we illustrate how we write pseudocode specifying the behavior of the roles that each node plays during execution of this protocol from algorithmic description We then convert this into another pseudocode that is suitable for nesC programming In the source role the node reads the data and forwards them to every neighboring node In the intermediate node role any data received are forwarded to all neighboring nodes In the sink node role the data received are processed and stored for future use Now from the behavior observed in these individual roles we can build the pseudocode for nesC programming This basically involves distributing the role behaviors acro
326. rithms Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 297 www finebook ir 298 SECURITY IN SENSOR NETWORKS 16 2 2 Communication Issues Standard sensor nodes equipped with low power radio transmitters commonly have a range of under 20 m In order for these nodes to communicate with a base station any significant distance away they need to use multihop routing Because the base station doesn t communicate directly with most of the nodes in its WSN it cannot efficiently manage key distribution for its nodes because of the high overhead that this would require Also security in traditional networks depends on well established protocols that assume that a reliable media exists for any kind of authentication to take place In sensor networks issues of latency conflicts and unreliability of the underlying media arise rendering most authentication schemes suboptimal for these classes of network 16 2 3 Hostile Environments Distributed sensor networks DSNs that are deployed in hostile environments must account for the fact that individual nodes can be easily captured A captured public key server could potentially be used to disclose a large number of keys compromising the network Finally the intrinsic properties of sensor networks distributed in nature and avoiding central management
327. rk Modern Sensing Technol special issue 90 72 86 2008 Johnson D B The Rice University Monarch Project Technical Report Rice Univ 2004 Johnson D S The NP completeness column An outgoing guide J Algorithms 6 434 451 1985 Kalidindi R V Parachuri S Basavaraju C Mallanda A Kulshrestha L Ray R Kannan and A Durresi Sub grid based key vector assignment A key pre distribution scheme for distributed sensor networks CWN 2004 Kalidindi R R Kannan S S Iyengar and A Durresi Sub grid based key vector assign ment A key pre distribution scheme for distributed sensor networks J Pervasive Comput Communi 2 1 35 43 March 2006 Karl H and A Willig Protocols and Architectures for Wireless Sensor Networks John Wiley amp Sons Inc 2005 Karlof C and D Wagner Secure routing in wireless sensor networks Attacks and counter measures Proc 1st Int IEEE Workshop on Sensor Network Protocols and Applications Univ California Berkeley 2003 Karsai G A Ledeczi J Sztipanovits G Peceli G Simon and T Kovacshazy An approach to self adaptive software based on supervisory control Proc Int Workshop on Self Adaptive Software 2001 Klein P N Efficient Parallel Algorithms for Planar Chordal and Interval Graphs Ph D thesis MIT Cambridge MA 1988 Klein P N Efficient parallel algorithms for chordal graphs Proc IEEE 29th Symp Foundation of Computer Section 1988 pp
328. rla Glomosim A library for parallel simulation of large scale wireless networks Proc Workshop on Parallel and Distributed Simulation 1998 pp 154 161 ZigBee Wireless Networking Newnes Publications 2008 www finebook ir Index Acquaintance Group Pattern 3 15 17 Active Object 4 17 Addressing Scheme 4 Algorithm 5 Alternate Bit Based ARQ Protocols 6 Application Layer 8 Application With Sensing 8 Asynchronous Commands 8 Atomicity Service 9 Automatic Repeat Request Protocol 9 Beaconing 9 Biconnected Graph 9 Binary Trees 12 Block 15 Block Graph 15 Carrier Sense Multiple Access Protocol 15 Chord 17 Circular List 21 Client Level 21 139 Clique 21 140 Clique Covering 22 Collaborative Processing 22 Collection Tree Protocol CTP 22 Collision 23 Coloring 23 Complete Binary Tree 23 Complete Graph 24 Connected Graph 24 Connectivity 24 Content Based Addressing 24 Contention Based Protocols 27 Correlation 27 Cycle 27 Data Aggregation 30 31 Data Routing 32 Datalink Reliability 33 Disconnect Graph 34 Distributed Sensor Network DSN 35 Doubly Linked List 36 Dynamic Routing 36 Dynamic Topology 37 Event Driven Programming 41 Explicit Embedding 45 Fanin Wiring 45 Fanout Wiring 46 Flooding 51 Free List 51 Full Binary Tree 51 Full Function Device 53 Gateway Device Properties 53 GlomoSIM 56 Graph 56 Graph Travers
329. rror 7 3 1 Sensing the Temperature In this application we show how we can read the temperature using a sensor node mote and turn an LED on SightC The module SightC uses five interfaces and thus it needs to provide im plementations for all the events that are declared in those interfaces Every function defined in this module implements an event except the last function errorReporter www finebook ir APPLICATIONS USING THE INTERFACE SplitControl 123 which is called locally to turn the red LED on off We need to find an implemen tation for each one of these interfaces We already know which system components provide implementations for Boot Timer and Leds Reading the temperature sensor will require device dependent command functions so let us commit ourselves to the TelosB mote Crossbow Technology The TinyOS system provides a component named Msp430InternalTemperatureC which provides an implementa tion for the Read interface We can directly wire to this component but for the sake of readability let us define a configuration called TemperatureC to provide the inter face Read This is done simply by exporting Msp430InternalTemperatureC Read as Read using the wiring operator Note that Read is declared in the statement provides in TemperatureC generic configuration TemperatureC provides interface Read lt uintl16_t gt implementation components new Msp430InternalTemperatureC Read Msp430InternalTempe
330. rst we set the current address of this mote call AMA setAddress sensorgroup sensoraddress if self source then read x 1 www finebook ir RUMOR ROUTING 187 readDone Note This is executed only by source node if self source then 1 d data from read operation for each randomly selected neighbour k send to node k data d event void AMSend sendDone message_t msg error_t err if err FAIL call Leds ledOOn receive if self sink thene xtract packet and process if self intermediate amp firstTime then firstTime false extract data for each randomly selected neighbour k send to node k data d event void AMSend sendDone message_t msg error_t err if err FAIL call Leds ledOOn J async event void AMA changed event void RadioControl stopDone errort err module Tracking uses interface Boot uses interface ReadStream as Temperature www finebook ir 188 TECHNIQUES FOR PROTOCOL PROGRAMMING uses interface Timer lt TMilli gt as Timer0 uses interface Timer lt TMilli gt as Timerl 9 14 TRACKING In this section we show how as imple program tracking utility can be written for a sensor network In this example it tracks the changes in temperature over time We now build the component for tracking by using the interfaces as shown configuration TrackingAppC imp
331. ruct the link layer packet repeat forever www finebook ir 162 TECHNIQUES FOR PROTOCOL PROGRAMMING transmit frame to receiver node j Can be done using Receive receive parallel 1 P1 wait for ack if ack positive exit repeat loop else continue it is negative ack waited long enough P2 if time out exit repeat loop end of parallel end of repeat sleep until woken up Receiver node j repeat for ever Can be done using Receive receive p receive packet result checksum test p if result success send success exit repeat loop else send failure end of repeat loop Sleep until woken up The receiver node sends a positive acknowledgment and goes to sleep if it receives uncorrupted data Otherwise the receiver after sending a negative acknowledgment waits for a retransmission of the data TRANSMITTER ARQ FSM Model Transmitter S0 frame getFrame www finebook ir ALTERNATING BIT BASED ARQ PROTOCOLS 163 transmit frame go to S1 81 receive if acktype ve then frame getFrame transmit frame else transmit frame upon ack go to s1 Use Receive receive for this global frame acktype ve booted frame getFrame transmit frame receive actype extract ack type if acktype ve then frame getFrame transmit frame else
332. ructure Our directory structure is divided as follows LSU SensorSimulator inc This has the header files of base classes src This has the base classes for all the layers CoordinatorBase LayerBase PhyLayerBase MacLayerBase and NetLayerBase www finebook ir APPENDIX 273 AppLayerBase RadioBase BatteryBase CPUBase WirelessChannelBase TargetNodeBase cConsumer samples This has the simple classes derived from base classes it also includes implementation of directed diffusion with GEAR and MAC 802 11 phy layer mac layer netw layer app layer hwmodels layer wireless ch common Each subdirectory has implementations of that layer Hardware models include battery CPU and radio modules These include simple hardware models Thecommon directory has CoOrdinator packet structures for network and MAC layers and other constants and attributes used for simulation This directory is derived by all other directories of the sample folder The wireless channel subdirectory has a simple wireless channel that introduces a delay Delay D is calculated as follows distance between 2 communicating nodes speed of light The class name has to be specified in the omnetpp ini file if the user wants to try the other implementations specified above To try a new implementation the following steps are recommended 1 Files Each layer has three basic files cc hand and nedfiles The user needs all three of these file
333. s PI Ps where P 1 2 3 P 2 3 4 5 6 Ps 5 6 Py 6 7 8 9 Ps 4 10 Let vi v v be a cy cle of a graph Any edge joining two nonconsecutive vertices Vi and V is called a chord 4 15 DEFINING DATA STRUCTURE OF SPANNING TREE PROTOCOLS In the following paragraphs we define a new characterization of the data structure focusing on the programming aspects of sensor networks To do this we define flooding replicated packet arrivals and neighboring nodes all in the context of spanning tree data structures Flooding In its simplest definition floading refers to a situation in which a networking device is overwhelmed with packets to such an extent that its pro cessing capabilities are severely degraded Most devices facing such a situation simply drop packets Neighboring Nodes The term neighboring nodes refers to sensor nodes in the same proximity as another node Typically sensors collaborate with neighbroring nodes in processing tasks ranging from intermediate routing to computational tasks with their direct neighbors More details on these networking concepts are provided in the chapters on wireless sensor programming e g Chapter 7 4 15 1 Flooding Owing to the broadcast nature of wireless sensor networks messages can be easily send to a node s neighbors then the neighboring nodes rebroadcast the packets to their neighbors within range Flooding generates replicated packet arrivals to each node 4 15 2
334. s robustness may be improved 10 3 1 Example While the architecture of the group can be modeled as a directed acyclic graph in this example we will assume that a hierarchical tree structure exists among the agents Each node has a parent node except the root node and a set of child nodes see Fig 10 2 The overall task of a subgroup formed by the nodes of a subtree is broken down into subtasks by the node at the root of the subtree distributed to the children the results from the children are then aggregated and finally sent to the parent Typical operation of a node in this hierarchy Given task T e g measure temperature and brightness while 1 decompose task T into subtasks T1 T2 Tn Assume n children distribute subtasks to children c1 cn sense x r receive results from children q compute function f x r communicate result q to parent 10 3 2 Types of Group Abstractions Abstractions using nodes can be defined based on geographical distribution of nodes tasks assigned to each node and specific types of capabilities that may exist across the nodes For example in abstractions determined by geographically constrained group GCG of nodes the nodes are placed within a chosen geographic area The role chosen for each node will depend on relationship between the nodes and the www finebook ir 204 SENSOR DEPLOYMENT ABSTRACTION Application Collaboration groups
335. s 214 INSPIRE Innovation in Sensor Programming Implementation for Real Time Environment 12 1 Motivation and Background 215 12 2 Software Microframework Requirements 236 References 237 Performance Analysis of Power Aware Algorithms 13 1 Introduction 239 13 2 Service Architecture 242 13 3 Approaches To WSN Programmability 248 13 4 Simulation Capabilities 249 13 5 Benchmarking 251 13 6 Conclusion 251 Problems 252 References 252 Modeling Sensor Networks Through Design and Simulation 14 1 Introduction 254 14 2 Why a New Simulator 254 14 3 Currently Available Simulators 255 14 4 Simulation Design 257 14 5 Implementation Details 261 14 6 Experimental Results 265 14 7 Final Comments 271 Appendix 272 Acknowledgments 275 Problems 275 References 275 MATLAB Simulation of Airport Baggage Handling System 15 1 Introduction 277 15 2 Background 277 15 3 Proposed Architecture 283 www finebook ir 207 215 239 253 277 CONTENTS xi 15 4 Simulation Results and Discussion 283 15 5 Source Code 286 Problems 295 References 296 16 Security in Sensor Networks 297 16 1 Introduction 297 16 2 Security Constraints 297 16 3 Denial of Service Attacks in Multiple Layers 298 16 4 Some Well Known Algorithms for Security Problems 302 16 5 Secure Information Routing 302 16 6 Security Protocols for Sensor Networks 303 16 7 Final Comments 303 Problems 303 References 304 17 Closing Comments 305 Bibliography 3
336. s ActiveMessageC components new TimerMilliC algorithm2C Boot MainC Boot algorithm2C Random gt RandomC algorithm2C Leds LedsC algorithm2C AMPacket gt AMSenderC algorithm2C AMSend gt AMSenderC algorithm2C AMA gt ActiveMessageAddressC algorithm2C PacketAcknowledgements gt AMSenderC algorithm2C Receive gt AMReceiverC algorithm2C RadioControl gt ActiveMessageC 8 4 2 Neighborhood Table Construction In this application we use the standard interfaces shown in the program above www finebook ir SYNCHRONIZATION IN WIRELESS SENSOR NETWORKS 141 Wiring for Neighborhood Table Construction The interfaces are implemented using the system provided components shown above The implementation details are discussed below implementation void sendSensorInformation uint 16_t void populateSensorTable uint 16_t uint 32_t int NO_OF_MOTES 30 sensorInfo sensor am_addr_t sensoraddress 0 uint32 t sensormac 0x001EDD32 Pre determined mac address of current sensor am_group_t sensorgroup 1 sensorData table NO_OF_MOTES int index 0 event void Boot booted call RadioControl start event void RadioControl startDone error_t err call AMA setAddress sensorgroup sensoraddress call TimerO startOneShot 6000 event void TimerO fired message t msg sensor sensor Info call AMSend getPayload amp msg sensor gt id sensoraddress call AMSend send AM BROAD
337. s for the implementation We are providing these three files for every layer with minimum functionality You can add your own code and just do make and run the simulation by changing the necessary configurable parameters in the omnetpp ini file in the samples directory 2 Procedure Move into samples directory This directory has subdirectories for each layer Each directory has examples of various implementations at each layer Also you can see the files New Layer cc New Layer h and New LayerModuleDefined ned for a new user to start using the simu lator You can add the various parameters of a module that you will be using in your implementation in the need file After adding the code or making changes www finebook ir 274 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION for the existing implementation do make and then go back to the samples directory The class name of this new implementation has to be speci fied in omnetpp ini so that the simulation considers your implementation for execution e g sensorNetwork Nodes strMACLayerType MAC 802 11 With this functionality the user can just add his her protocol at that particular layer with all other protocol layers being the same 3 Building Simulations This section explains the basic concepts behind the Sen sorSimulator The class hierarchy is explained and all relevant functions of the base modules are introduced Detailed description is also available in the AP
338. se Typically it is not feasible to associate mass stor age devices at the level of a sensor and the amount of memory available in a resource constrained sensor is limited Thus it is necessary to manage the data in a DSN and effectively synthesize information that is useful for decisionmaking Data management considerations for periodic systems are more critical because of issues of data freshness The computing subsystem must support multiple system inter faces to effectively integrate with other systems For the interface with the physical environment it is necessary to interface to proprietary and open sensor interface standards For example several sensors interface with Ethernet or SERCOS To al low users to work with emerging pervasive devices or to incorporate the DSN as an infrastructure for a smart space for automation 4 the DSN must support open interfaces that are based on XML eXtensible Markup Language or such other technologies The implementation of these functions is discussed under the categories of process ing architecture distributed services and sensor operating systems 7 2 Distributed services facilitate the coding and operation of a DSN and are provided by a dis tributed operating system that is represented by the collection of operating systems on each sensor Transparency refers to the ability to regard the distributed system as a single computer Tannenbaum 6 defines several forms of transparency for distribut
339. sensor networks Contrast them with the char acteristics of computer networks e g the Internet 2 8 Choose your favorite real world wireless sensor network application and list five causes sources of its unpredictability 2 9 Research and prepare a two page report describing a dynamic routing strategy for a wireless sensor network Discuss its strengths and weaknesses in detail 2 10 Describe a real world heterogeneous wireless sensor network application 2 11 Discuss the data management issues in wireless sensor networks 2 12 Consider the habitat monitoring application discussed in the text see also Fig 2 1 For this application consider the scenario where the sensors are dropped from an aircraft over a large area on the ground periodically An important aspect in such scenarios relates to the management of the sensors and the resulting network Compare the resulting network characteristics with the standard computer network characteristics particularly with respect to the following attributes discussed in Section 2 2 a The number of nodes in the network b Node failure c Communication across nodes d Density of deployment e Topology f Node IDs 2 13 An interesting situation arises when sensors are mounted on tiny moving objects such as robots obviously increasing the cost of the network For the habitat monitoring application mentioned above comment on the network characteristics with respect to the following issues
340. sor of computer science in the College of Engineering and Science at Louisiana Tech University USA He holds the W W Chew Endowed Professorship at Louisiana Tech and directs the Center for Secure Cyberspace He has won various distinctions including ACM Distinguished Scientist 2008 research commemoration awards at Louisiana Tech University 2002 2006 2007 2008 outstanding research faculty and faculty Circle of Excellence Award at Northeastern State University Oklahoma and as a student was awarded the President s Gold medal for Academic Distinction Professor Phoha holds an M S and a Ph D in Computer Science from Texas Tech University He has done fundamental and applied work in anomaly detection in network systems in particular in the detection of rare events His current research interests include autonomy and security issues in sensor networks He has eight patent applications and many reports of inventions He is author of over 90 publications and author editor of three books Internet Security Dictionary Springer Verlag 2002 Foundations of Wavelet Networks and Applications CRC Press Chapman Hall 2002 Quantitative Measure for Discrete Event Supervisory Control Springer 2005 Dr N Balakrishnan N Balakrishnan is a scientist of high international repute and is well decorated with prestigious awards He received his B E Hons in Electronics and Communication from the University of Madras in 1972 and Ph D from the Indi
341. sors and a wireless sensor network that can be ported to different classrooms and laboratories NesC provides a C type component based language In NesC a module is the lowest level of component abstraction that implements any commands provided in its interface It may directly address a particular hardware component such as a light sensor providing methods that abstract the actual operation of that particular hardware component Several modules may be grouped together using a configuration to form a larger component 18 ORGANIZATION OF THE BOOK The book is organized as follows In Part I we present an overview of the subject of sensor network programming beginning with a general introduction in the remainder of this chapter Chapter 1 Chapter 2 gives a general description of the wireless www finebook ir ORGANIZATION OF THE BOOK 11 sensors It explains the basic components of a sensor its sensing environment and the various roles that a sensor can play in a wireless sensor network Chapter 3 discusses current sensor technology including the major families and types of sensors currently in use including the Mica Telos Tmote Sky families and others Part II provides a general background for sensor network SN programming beginning with discussions on data structures for sensor computing programming in Chapter 4 Sensor computing programming of individual sensors in a SN environment requires an understanding of data structures such
342. ss the commands booted readDone and receive where a node chooses a role initially to guide the execution of the nesC pseudo code 9 13 RUMOR ROUTING In rumor routing which is somewhat similar to flooding a node X has an event e in which other sensors may be interested In order to achieve this goal X lets www finebook ir RUMOR ROUTING 185 some of its neighbors A know about the event Note In flooding all neighbors are informed about the event Now node K forwards the data about the event to some of its neighbors but at the same time records in its memory that X is its neighbor from where event e originated This information is necessary if some node wants to reach X later In the Fig 9 3 we thus see that X chooses to inform its neighbors A and E about the event ignoring other neighbors At this point some of the nodes at hop distance 1 know about the event and that there is a route to the node X The routing thus continues for some further hop distance and terminates in this example at hop distance 4 where nodes D and G also know about the event at X 9 13 1 Example In the pseudocode below each node chooses a few nodes randomly from its neighbors We have assumed that the node X goes to sleep after transmitting its data whereas the other nodes are still a wake performing other activities Unlike in the scenario flooding rumor routing is followed by the next phase where other nodes want to send queries to nod
343. st number 0 debug level 10 EVB LED1 OFF EVB_LED2_OFF get this for reference will use the LSB as srcEP for indirect message halGet ProcessorIEEEAddress amp myLongAddress 0 ifdef LRWPAN_COORDINATOR aplFormNetwork wait for finish while apsBusy apsFSM EVB_LED1_ON conPrintROMString Nwk_formed n else do www finebook ir 222 INSPIRE aplJoinNetwork while apsBusy apsFSM wait for finish if aplGetStatus LRWPAN STATUS SUCCESS EVB LED1 ON conPrintROMString Network EREE ENE E EEE EEEN E ac uctus oe Join _succeeded n conPrintROMString My Sa a ater nde st Re ec in aa qS pes ShortAddress is conPrintUINT16 aplGetMyShortAddress conPCRLF conPrintROMString Parent _LADDR conPrintLADDR aplGetParentLongAddress conPrintROMString Parent SADDR conPrintUINT16 aplGetParentShortAddress conPCRLF break else conPrintROMString Network_Join_FAILED eee er cre e rca inta es a Cispa erate ere cua a Waiting then trying again n my_timer halGetMACTimer wait for 2 seconds while halMACTimerNowDelta my_timer lt MSECS TOMACTICKS 2 1000 while 1 endif ifdef LRWPAN RFD announce ourselves to the coordinator so that we can test indirect messaging this is only necessary if there are routers between us and the c
344. state should the atomic operation fail Typically atomicity is more important at the level of information based transactions and less important at the level of periodic data gathering The order in which data from various sensors are gathered and the nature of in teractions among the multiple sensors depends on the synchronization method The event service allows a sensor to register an interest in particular events and to be notified when they occur The time service is used to provide a systemwide notion of time An important application of system time is in the diagnostic function where it is used to establish event causality Two forms of time are possible clock time and logical time Providing a systemwide clock time that is globally known within a spec ified accuracy to all the controllers in a distributed system can be difficult Clock time can represent a standard Coordinated Universal Time UTC or it can be a common time local to the system Two common techniques are 1 to provide a hierarchical master slave system in which the time in the master sensor device is transmitted to the other slave sensors or 2 use a peer to peer distributed mechanism to exchange local times among various sensors For certain applications where affordable it is possible to use global positioning system devices as master clocks at the master sen sor devices to synchronize multiple controllers at the slave devices with the UTC Logical time pro
345. stimated by assuming either a linear model or a discharge rate dependent model In the linear discharge model the remaining capacity is cse f I t At At where Cj is the initial capacity of the battery and Z t is the current drawn by the sensor node in duration At This model assumes that there are no self discharges and that the battery does not deteriorate with age The discharge rate dependent model assumes that higher discharge rates effectively reduce the remaining capacity of the battery To allow various models to be implemented with the type of application we have a basic battery module BatteryBase which forms the abstract class for the different battery models BatteryLinear www finebook ir 260 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION is a subclass of BatteryBase and updates the energy depending on the number of consumers and the state of activity of the consumers BatteryDischargeRate is a subclass of BatteryBase and the energy consumption is a linear function of current 2 CPU Model The nodes in a sensor network are usually equipped with very low end processors or microcontrollers The power consumption for performing various operations should be very low and we have used a standard set of parameters for energy consumption by the processor model The processor needs different levels of energy consumption in the idle sleep and active states The processor power consumption model is very important and i
346. sumption after simulation ends 10 queries 900000 800000 7 700000 600000 500000 a SensorSimulator 400000 ns2 300000 200000 100000 0 T T T T 100 200 300 500 1000 2000 Number of Nodes Memory Kbytes FIGURE 14 12 Memory consumption before simulation starts 100 queries 14 6 3 Directed Diffusion with IEEE 802 11 MAC This series of experiments use MAC 802 11b at the MAC layer and compare the performance of sensor simulator with that of ns2 A simple pass through the physical layer is considered The simulation is run for 100 500 1000 and 2000 nodes The 600000 500000 400000 300000 SensorSimulator s ns2 200000 Memory Kbyt 100000 Number of Nodes FIGURE 14 13 Memory consumption after simulation ends 100 queries www finebook ir 270 MODELING SENSOR NETWORKS THROUGH DESIGN AND SIMULATION 3000 2500 7 2000 SensorSimulator 1500 ns2 Time sec 1000 500 _ __ 0 T T 100 500 1000 2000 Number of Nodes FIGURE 14 14 Comparison with ns2 execution time nodes in the sensor network are deployed randomly in various locations with the network size being configurable in the omnetpp ini file Figure 14 14 shows the relative performance The setup for the nodes is as follows Number of queries 10 Simulation time 300 s Network dimension varies with the numb
347. t emitting diodes on the sensors for debugging TOSSIM allows detailed simulation of TinyOS network stacks at the bit level therefore allowing both low and high level applications that require fine grained controls to be simulated TinyViz a graphical user interface GUI for TOSSIM can be used to visualize and interact with running TOSSIM simulations The TOSSIM simulator can mimic thousands of nodes and also has the ability to print debug information such as variables We make no pretense of providing in this book a reference manual for nesC or of TinyOS and students writing more than trivial programs will do well to avail themselves of specific reference sources on nesC and TinyOS Audience The book is suitable for upper division junior or senior undergraduate level or first year graduate level students We have presented the material in this book assuming that the reader is knwoledegeable in some conventional programming language and has basic familiarity with an operating system Although the programming language used in examples is nesC running under the TinyOS operating system in some cases particularly in Part III of the book further knowledge of sensors and networks will be helpful We have tried to arrange the material in an order that will facilitate following the chapters linearly However readers who are confident in the programming data structures have a basic understanding of sensors and want to focus more on software
348. t is assigned the serial number 1 The nodes at level 1 are numbered as 2 and 3 from left to right After assignment of serial numbers for all the nodes at level 1 the consecutive numbers are assigned for all the nodes at level i from left to right In Fig 4 25 a full binary tree of height 3 and the serial numbering of its vertices are given The following theorem can be easily verified for any full binary tree Theorem 4 2 In a full binary tree that is serially numbered 1 The left child of node i is 2i 2 The right child of node i is 2i 1 3 The parent of i is i 2 In a full binary tree if some highest numbered vertices are removed it is called a complete binary tree Figure 4 26a shows a complete binary tree Fig 4 26b a full binary tree and Fig 4 26c an incomplete binary tree In a complete binary tree the numbering is continuous and coincides with the numbering of the full binary tree of the same height Theorem 4 2 holds for any complete binary tree Consider a binary tree in which every pendant vertex is assigned a positive weight Let v and w denote the level number and weight of a pendant vertex respectively Consider the sum v1 v2 where the sum is taken over all pendant vertices This sum is called the weighted pathlength For example consider the binary tree shown in Fig 4 27 It has six pendant vertices d q g p r f They are given weights 5 10 7 3 4 8 respectively The weights and the level numbers are shown
349. the presence of the transmitter The pseudocode describing how beaconing works is presented below begin while true broadcast address random time Sleep random time end As we did before we assume that the node has the three basic capabilities sense broadcast and sleep as enumerated above In this example we use a procedural style of code Procedural Beaconing begin senseMedium signal if signal equals absent return no signal elif signal equals weak return medium busy or no data elif signal equals collision return collision elif signal equals strong return strong end Sometimes certain routing schemes may employ the use of lists to keep track of their nearest neighbors A sample pseudocode and associated nesC configuration showing how a neighborhood table construction can be done is shown below www finebook ir 140 ALGORITHMS FOR WIRELESS SENSOR NETWORKS initialize table table i 0 tl time now while i n data packet sense data address address of data packet if not present address table add address table i if time now gt tl max time break define MOTE_AM_ID 10 configuration algorithm2Appc implementation components MainC components RandomC components LedsC components new AMSenderC MOTE AM ID components new AMReceiverC MOTE AM ID ActiveMessageAddressC components ActiveMessageAddressC components algorithm2C component
350. the scheduler encapsulates all the details of thread safe event exchange and queuing As sensor programming is network centric we need to combine a reliable micro kernel OS to build a framework for network embedded OS The main design goals are to design an adaptive MAC layer that needs to be interfaceable to many radio stacks and have dedicated tasks that sense periodic data from physical sensors in the background Thus we design a sensor mote with two microcontroller units MCUs Dividing the functionality of sensing and communication can be accomplished with its own control The requirements are easy programmability of the sensors and its control for any deployment To have a microframework we design a state machine object that allows the creation of active objects that can then send messages without blocking using a priority queue This microframework can be extended to support mi crosecond accuracy using specific hardware abstraction layers HALs for available manufacturer sensor boards There are several choices for programming the Texas Instruments TI MSP430 Here we try to abstract the TI architecture using a coresi dent small footprint OS that allows the kernel booting into a mode f to have complete real time control and at the same time be compatible with all the sensor interfaces that the MCU can control and configure This allows the writing of sensing applications by creating tasks and timers and scheduling individual tasks on the part
351. ticulation point G itself is the only biconnected component 2 For every articulation point as V let Vi V2 V3 V denote the sets of vertices in the different components of G a 3 Let G graph induced by V Ua i 1 2 3 5 These definitions and procedures can be generalized for connected components k gt 2 A block graph can be represented by its block cut vertex tree representation BC tree For example consider the block graph shown in Fig 4 40a Its blocks www finebook ir 80 DATA STRUCTURES FOR SENSOR COMPUTING FIGURE 4 39 a A graph G that is connected but not biconnected b biconnected com ponents of G are Bj 1 2 B5 2 3 4 5 B4 5 6 7 B4 5 9 Bs 5 8 Bg 8 106 B5 10 11 Bg 10 12 13 Bo 4 14 The cut vertices are 2 4 5 8 10 The vertex set of the BC tree consists of the blocks and the cut vertices A block and a cut vertex are joined by an edge only when the vertex is contained in the block The BC tree of the block graph depicted in Fig 4 40a is given in Fig 4 40b FIGURE 4 40 a Block graph b the BC tree of a block graph www finebook ir PLANAR GRAPHS 81 4 11 PLANAR GRAPHS In this section we define the graph G as a pair of two sets V and E where V is any nonempty set and V is a subset of V x V Any graph can be schematically represented in more than one way Figure 4 41 is a representation of the graph G V E where v a b c d e f
352. tivity of at least k is said to be k connected where k is a positive integer The following results immediately ensure 1 A connected graph is biconnected if and only if for any two vertices u and v there are at least two edge disjoint paths from u to v 2 A connected graph is triconnected if and only if for two vertices u and v there are at least three edge disjoint paths from u to v 3 A connected graph is k connected if and only if for any two vertices u and v there are at least k edge disjoint paths from u to v The graphs shown in Fig 4 36a and 4 37 are connected but not biconnected Any cycle C is biconnected Figure 4 38 shows a triconnected graph If d denotes the minimum degree of graph G vertex connectivity of G is at most d We have already seen that a disconnected graph has more than one connected component a connected component is a maximal connected subgraph Similarly we are defining the biconnected component as a maximal biconnected subgraph of the graph If the graph itself is biconnected there is only one biconnected component Otherwise the biconnected components can be found Figure 4 39b shows the bicon nected components of the graph shown in Fig 4 39a The biconnected components are also called blocks A graph G in which every block is a complete subgraph is called a block graph Figure 4 40a shows a block graph The biconnected components of a graph G can be found by the following operation 1 If G has no ar
353. tomicity does not necessarily mean total absence of preemption www finebook ir A SIMPLE PROGRAM 107 6 2 5 Wiring In our earlier example we explained how to perform wiring so that module commands and events can be appropriately called during execution There are a few more special cases that still need discussion configuration CC2420TransmitC provides interface Init implementation components Alarm CC2420TransmitP Init Alarm Init CC2420TransmitP Fanout In Fig 6 2 we have shown four components CC2420TransmitC CC2420TransmitP Init Alarm and TransmitApp corresponding to configu ration CC2420TransmitC shown using nesC code above The Transmit App component makes use of the Init interface provided by CC2420TransmitC TransmitApp FIGURE 6 2 Flowchart illustrating the concept of fanout wiring www finebook ir 108 PROGRAMMING IN NesC This Init interface is wired to the interfaces provided by components Alarm and CC2420TransmitP Therefore when a command is called from TransmitApp to the Init interface it fires the respective commands in both Alarm and CC2420TransmitP components This multiple wiring from one component to multiple components is called Fanout wiring Correspondingly there is the opposite situation where multiple callers are calling one callee as in the example below which is known as fanin wiring Fanin Suppose that a command c is invoked from component A and
354. tus at Stan ford University for his fundamental contributions to the programming in Computer Science Professors Daniel Siewiorek Carnegie Mellon John Hopcroft Cornell University Juris Hartmanis Cornell University Thomas Kailath Stanford University and K Mani Chandy Cal Tech have all inspired the authors for coming up with the first book on sensor programming Also Dr S S Iyengar would also like to dedicate this book to all his former and current Ph D students his son Vijeth Iyengar and finally grandson Ranvir Iyengar Professor Phoha would like to dedicate this book to Shiela Rekha Krishan and Vivek S S Iyengar N Parameshwaran Vir Phoha N Balakrishanan Chucka Okaye www finebook ir Contents PREFACE FOREWORD ACKNOWLEDGMENTS ABOUT THE AUTHORS NOTATIONS AND ABBREVIATIONS I OVERVIEW 1 Introduction 1 1 Some Foundational Information 3 1 2 Next Generation Sensor Networked Tiny Devices 1 3 Sensor Network Software 6 1 4 Performance Driven Network Software Programming 8 1 5 Unique Characteristics of Programming Environments for Sensor Networks 10 1 6 Goals ofthe Book 10 1 7 Why TinyOS and NesC 10 1 8 Organization of the Book 10 1 9 Future Demands on Sensor Based Software Problems 12 References 14 2 Wireless Sensor Networks 2 1 Sensor Network Applications 17 2 2 Characteristics of Sensor Networks 20 2 3 Nature of Data in Sensor Networks 24 Problems 24 References 25 3 Se
355. twork and Routing Layer Neglect Greed Homing Misdirection Black Holes Data Link MAC Layer Collision Exhaustion Unfairness Physical Layer Jamming Tampering FIGURE 16 1 An illustration showing possible attack types on each layer frequences Jamming can be eliminated by changing frequencies or by using a wide band of frequencies via modern spread spectrum techniques Figure 16 2 illustrates how a jammer could operate in a wireless sensor network Tampering Tampering is when the network hardware is physically attacked dam aged or otherwise compromised The best defense against tampering is hiding the nodes or making them physically resistant to attack 16 3 2 Datalink Layer A collision occurs when two packets are sent on the same channel at the same time corrupting the recieved data Because radio is inherently a half duplex system a node cannot easily receive during transmission in order to check for collisions Further more even if this were possible the signal received at a node s neighbors would be significantly different from that at point blank range So in wireless communications Jammer FIGURE 16 2 An illustration of a jammer in a sensor network www finebook ir 300 SECURITY IN SENSOR NETWORKS a collision appears for all intents and purposes as corrupted data at the reciever s end Unfortunately if the attacker has already compromised one or more nodes and is able to send malicious packets onto the
356. ulator that is designed for use as a sensor network simulator must be able to simulate the larger number of nodes 2 Propagation models When the user simulates a large sensor network the topology used might be that of an actual deployment It is therefore very important that the wireless channel be modeled accurately The simulation can then be used to detect connectivity and other such problems 13 4 SIMULATION CAPABILITIES 13 4 1 Unmodified Code Simulation One primary focus of this implementation is to allow FARMS applications to be simulated inside GlomoSIM Forcing the programmer to make GlomoSIM specific 3 modifications in the code decreases from the ease of use of the simulator as shown www finebook ir OVI8 VINSO OVWEVWSD OVWE VASO T T esueq ugs oBuel olpey uoisnyig untpelN Qq3ads sedg Hovi Es 9 208 333i zl SVA 8 voNWSO NEN OL S oz Sc o SE Q JUIS o oinos Wold spunol Bunnoy egue1 orpe1 juvjsuoo yum jueurko dop Arsuop usrq pue umrpour esreds 107 o ueuuogied DYN 2 urs o 2mos woy ooueunoj1ed oooloid doun nui q peeurojsnpo je ooueunj1ed ooojo1d uormeSoi133e vjep e2o e Aqgore1erg Sunojsn o oAndepe ASrouo Ao HOVAT pue uorsnjjip pojoourp oo1Aop uonen eAo pue SuroeurSuo Suruue d surojs s qrHHdS SUIN sunpuos e dounnur pue uonesoisse vjep 10 uosireduioo o ueuuogiod OVIN prer TAASIA
357. untime abstraction is that all the infrastructure used by the kernel is simulated to provide wireless communications using renewable energy resources with its unique extended lifetime model This allows it to scale all the code to any processor The implementation of INSPIRE on a target prototype node occupies less than 10 40 kB www finebook ir SENSOR NETWORK SOFTWARE 7 kilobytes of code memory for details refer to Chapter 10 The distributed source coding implementation is used to measure the sensing activity and memory overheads using traditional sensor applications without constraints but more importantly we highlight the reliability of the transmitted data from the measuring applications 13 1 Technology Driven Software Individual updates of software are impractical because of the large number of nodes and the relative inaccessibility of deployed nodes One solution for updating software in sensor nodes is the deployment of a support network of small mobile temporarily attachable nodes with virtual connections from a host PC to individual nodes This scheme allows the use of standard tools to update the software in the individual sensor nodes For many sensor networks in field applications such as sensors deployed in unreachable places such as in water or trees it is desirable to remotely update the software on the sensor nodes The following are a few issues to be considered when updating the nodes with software updates Up
358. ustration of collaborative processing among nodes 47 GRAPH AND TREES One of the most powerful ways to foster dynamical event tracking applications is to use an embedded data structure such as a universal data structure This data structure helps in developing a source of conceptual integrity from a programming perspective This captures the dynamic connectivity of the sensor network at each node see Fig 4 14 As each node can be added to or deleted from the network in real time it would need to have to have a unique address and information on its relative position This mechanism allows us to dynamically maintain the node addresses consistently over time This section introduces an important data structure the graph model 3 1 and explains various types of graphs and some properties that are used in later chapters 4 7 1 Preliminaries Let V be any set Let be a subset of V x V The pair V E is called a graph We denote G V E V is called the vertex set and E the edge set The element of V are called vertices and the elements of E are called edges Let V a b c d e f and E a b b c cd ce a d b d The graph G V E can be represented by Fig 4 15 If e vi v2 is an edge then we say that e incidents on v and v In such case v and V are said to be adjacent vertices An edge v v is called a self edge If ei v v2 and e2 vi v2 then e and e are said to be parallel edges A graph having no p
359. vailable resources It is for this reason that we discuss how a TinyOS based programming environment can be set up for both Linux and Windows operating systems 3 4 4 Installing TinyOS in Linux Using the very popular Ubuntu operating system these guidelines explain how TinyOS can be installed and configured in a minimal number of steps In admin istrator mode perform the following steps 1 In your fresh installation of Ubuntu open the file sources list located in etc apt with your favorite text editor 2 Add the TinyOS repository to your sources list file deb http tinyos stanford edu tinyos dists ubuntu hardy main 3 Run the aptitude package management program to update local repositories sudo apt get update 4 Ensure that all necessary tools required are already installed sudo apt get install build essential 5 Install TinyOS sudo apt get install tinyos 2 1 0 www finebook ir PROBLEMS 37 6 On installing TinyOS a few environment ralated variables have to be set In a Final step edit your bashrc file stored in your home directory and add the following lines export TOSROOT opt tinyos 2 1 0 export TOSDIR TOSROOT tos export CLASSPATH STOSROOT support sdk java tinyos jar export MAKERULES TOSROOT support make Makerules export PATH opt msp430 bin PATH Your TinyOS installation and configuration are now complete You can run the check env command to view system sanity information 3 4 2
360. ved transmitter A Let Nmax 4 repeat forever load buffer 0 7 0 packetO 1 packetl 7 packet7 i 1 j is number of packets for which acks received go 1 PAR P1 while i lt 7 shared if go i i 41 send i packet go 1 if i 7 wait P2 wait for ack Receive receive should solve this problem j extract control digit ack delete j packet j from the buffer shared i j 1 go 0 for ever www finebook ir ALTERNATING BIT BASED ARQ PROTOCOLS 167 receiver B while 1 receive from node a packet p Receive receive contorl digit j extract control digit from packet p send ack control digit data extract data from packet p process data Transmitter A GENERALIZED VERSION Transmitter s0 c 0 k 0 goto s1 sl while k lt 7 send a k k s2 receive cbit extract cbit if cbit c then c c 1 mod 8 else k c go to s1 nesC code booted c 0 k 0 transmit transmit while k lt 7 senda k Kk J receive cbit extract cbit if cbit c then c c 1 mod 8 else k c call transmit Receiver d GENERALIZED VERSION S0 receive from node a packet p contorl digit j extract control digit from packet p send ack control digit data extract data from packet p process data www finebook ir 168
361. ver the situation is quite different There are no such dominant protocols or algorithms and there will unlikely be any because a sensor network is often tailored for a particular application with specific features and it is unlikely that a single algorithm can always be the optimal one under various circumstances Many other publicly available network simulators such as JavaSims SSFNet and Global Mobile Information System Simulator GloMoSim and its descendant Quaint attempted to address problems that were left unsolved by ns2 Among them JavaSIM developers realized the drawback of object oriented design and tried to attack this problem by building a component oriented architecture However they chose Java as the simulation language inevitably sacrificing the efficiency of the simulation SSFNet and GloMoSim designers were more concerned about parallel simulation with the latter more focused on wireless networks They are not superior to ns2 in terms of design and extensibility The design of wireless sensor networks requires us to simultaneously consider the effects of several factors such as energy efficiency fault tolerance QoS demands synchronization scheduling strategies system topology communication and coor dination protocols This chapter presents the structural design of a new simulator for wireless sensor networks that is based on the discrete event simulation 9 16 framework OMNeT and results that demonstrate that the
362. verview of these security algorithms For more details and information on related projects refer to the references listed at the end of this chapter 16 4 4 KKID Sub Grid Based Key Vector Assignment A Key Predistribution Scheme for Sensor Networks In general a secure sensor network framework is of utmost importance as these vary sensors are placed in environments that pose a high risk of sensor capture and perhaps destruction In addressing this issue the employment of certain preventive mechanisms such as trusted third party authentication and public key systems render useless as these mechanisms oftentimes exhibit sub optimal resource requirements Key predistribution was introduced in 2003 2 to solve this problem Our scheme achieves connectivity identical to that of random key predistribution 2 but fewer using preloaded keys in each sensor node The design of our scheme is motivated by the observation that at present most key predistribution schemes employ random mechanisms that use a large number of keys and are unsuitable for sensor networks In this algorithm we extend the deterministic key predistribution scheme that we proposed in our earlierwork 3 which is based on assigning keys to sensors by placing them on a grid This approach has been further modified to use multiple mappings of keys to nodes In each mapping every node receives a distinct set of keys that it shares with different nodes The key assignment is done such that th
363. vides only the relative order of events in the system not their absolute clock time For many applications exact time may not be as important as ensuring that actions occur in the correct sequence or with in certain relative time intervals between events Many algorithms can be rewritten to use logical time instead of clock time to perform their function Providing logical clocks in a distributed system may be more cost effective if the applications can be restructured The management of shared resources across the network is supported through mechanisms that implement mutual exclusion schemes for concurrent access to resources All tasks in a sensor execute in an environment provided by the sensor operating system This operating system provides services to manage resources handle interrupts and schedule tasks for execution The operating system is said to provide real time services if the length of time required for performing tasks is bounded and predictable The operating sys tem is said to be non real time if such services are not supported Real time services are supported by providing either a periodic execution model or a real time scheduler e g rate monotonic scheduling These schedulers are priority based and can be pre emptive interruptible or not Preemptive scheduling can provide the fastest response times but there is an additional context swap overhead Depending on the way in which the scheduler operates the methods used to cod
364. vilian applications such as habitat monitoring environment observation and forecasting health applications vehicle traffic management and smart environments 2 1 4 Sensors Habitat monitoring is regarded as a driver application of wireless sensor networks and benefits the scientific community and facilitates ecological protection The use of sensor networks eliminates the potential impact of human presence and enables data collection at scales and resolutions that are difficult to achieve through traditional instrumentation A notable research project is monitoring seabirds on Great Duck Island in Frenchboro Maine a project conducted by the researchers from the University of California Intel Research Fig 2 1 The UC Berkeley Mica motes are deployed to collect data for studying seabird behavior during breeding seasons effects on the environment of nesting seabirds and related phenomena Research on tracking and controlling animals has also been conducted Wireless sensor networks have also been introduced for monitoring plants PODS is aresearch project at the University of Hawaii to test ecological environments with rare and endangered species The puspose of this research is to determine why endangered species of plants grow in one place but not in neighboring areas Instead of managing sensor networks in untraversed places remotely researchers from Intel Research deployed sensors in a vineyard and explored the future usage of sensor
365. void RadioControl stopDone error_t err event message_t Receive receive message_t msg void payload uint8 t len am_addr_t broadcasted_addr broadcasted_addr am_addr_t payload call Leds set 0 Clear debug leds if broadcasted_addr call AMA amAddress Generate a new addr call Leds ledoOn A collision has occured generateAddress call AMA setAddress www finebook ir DISTRIBUTED ASSIGNMENT OF NETWORKWIDE ADDRESSES 173 my_group my address broadcastAddress Only broadcast new address after a collision else call Leds led1On No collison yet return msg In implementing this scheme we use as shown in the program presented above a set of interfaces to achieve booting of the component random number genera tion sending and receiving messages with the neighbors and a timer to initiate the first broadcasting of a randomly generated address We have also shown above which implementation has been chosen for each interface In particular the interface RadioControl i e SplitControl has been implemented using the ActiveMessageC component define MOTE_AM_ID 10 configuration algorithm2AppC implementation 1 components MainC components RandomC components LedsC components new AMSenderC MOTE AM ID components new AMReceiverC MOTE_AM ID components ActiveMessageAddressC components algorithm2C components ActiveMessageC compone
366. where the randomly generated address is stored before being broadcast to the neighbors www finebook ir DISTRIBUTED ASSIGNMENT OF NETWORKWIDE ADDRESSES 175 As soon as the component is booted the component is turned on by RadioControl start where we perform the following tasks We first gen erate a random address and then place it in the global variable my address so that it can be used by other functions in the component We then check whether any other node has the same address as ours address collision and this is done by initiating a one shot monoshot timer that fires after some delay 6000 ms in our program If no such collision is observed we set the generated address as our address and we are done Our remaining task is only to reply to others queries regarding whether there are any address clashes event void TimerO fired message t msg am addr t temp temp am_addr_t call AMSend getPayload amp msg storing the address of my address in temp temp my address call AMSend send 0 x ffff amp msg sizeof amaddr t event void AMSend sendDone message_t msg error_t err if err FAIL Tf send fails switch on red led call Leds ledOOn event message t Receive receive message_t msg void payload uint8 t len 1 am addr t broadcasted_addr broadcasted_addr am_addr_t payload call Leds set 0 Clear debug leds if broadcasted_addr call AMA amA
367. wireless sensor networks Suppose that we augment the protocols with flexible dialog features where each sensor node engages intelligently with the other sensor nodes What will be the advantages Discuss the resulting overhead www finebook ir 14 1 17 1 18 1 19 INTRODUCTION Suppose that we view a WSN as a multiagent system MAS Suggest an application where this view will be appropriate What will be the undesirable aspects inherent in such a MAS model Traditional network systems are designed to satisfy strict specifications Be cause of the dynamic and uncertain environments in which WSN is employed traditional approaches may not be appropriate Investigate why this may be the case If self autonomy is one possible solution how can it help the sensor network in satisfying the application requierements What additional compli cations will this solution create for the application Indentify some aspects of security issues that are unique to WSN but may not be present in the traditional computer network systems REFERENCES 1 R R Brooks and S S Iyengar Multi Sensor Fusion Prentice Hall Englewood Cliffs NJ 1997 2 K Chakrabarty and S S Iyengar Scalable Infrastructure for Distributed Sensor Networks Springer Verlag 2005 3 S S Iyengar and R R Brooks eds Distributed Sensor Networks CRC Press Dec 2004 4 I F Akyildiz W Su Y Sankarasubramaniam and E Cayirci A survey of sensor networks
368. wireless sensor networks and provide accompanying pseudocode to explain the concepts Part IV presents real world scenarios in sensor network programming In Chap ter 10 we discuss some programming abstractions that simplify the development and deployment of sensors Chapter 11 presents standards for building WSN ap plications with a brief overview of the ZigBee networking standard Chapter 12 discusses an active sensor approach to distributed algorithms widely known as INSPIRE innovation in sensor programming implementation for real time envi ronments Chapter 13 explores the performance analysis of networks in some detail with respect to power aware algorithms Chapter 14 describes sensor network mod eling through design and simulation This chapter presents an architecture of a sensor simulator and a sensor node that is used in the simulator and further elaborates that OMNeT is a viable discrete event simulation framework for studying both the networking aspects and the distributed computing aspects of sensor networks We present the architecture of a sensor node that is used in the simulator and the general architecture of the simulator Chapter 15 presents a MATLAB implementation of simple data processing and decisionmaking logic to be used to detect and respond to events in an airport baggage handling system Chapter 16 consists of closing comments www finebook ir 12 1 9 INTRODUCTION FUTURE DEMANDS ON SENSOR BASED SOFTWARE
369. with packets and this can be done using the interface Packet We would also like to have the turn on and turn off control on the transmission receiving hardware components so we will employ the interface SplitControl for this purpose Module PacketSenderC shows nesC code that implements the packet sender Note that SplitControl has been renamed as PSControl Q fff node A node B FIGURE 7 4 Node A sending a packet to Node B www finebook ir 126 SENSOR PROGRAMMING PSControl start PSControl startDone H error success error fail Timer0 startPeriodic PSControl start TimerO fired amp SetLedBits SendPacket amp Leds set AMSend send Packet getPayload AMSend sendDone FIGURE 7 5 Control flow PacketSenderC SetLedBits Receive receive PN SetLedBits Leds set The module shows the list of interfaces used a few global variables declarations and the prototypes of the functions that are defined in the module In order to understand how the module works let us trace the control flow of execution starting from the event function booted from the interface Boot See Fig 7 5 where we have depicted the call sequence of the command and even functions as the module starts execution Solid arrows show one thread of control and the dotted arrow the other Let us start with the top most node in the graph namely booted As we may recall the TinyOS system be
370. work dispatches any available event to the appropriate state machine that handles the event and returns quickly to the main loop without ever waiting for events internally A variation of event driven programming is shown in Fig 4 1c for battery operated constrained wireless sensors Here the data are transmitted periodically to a sensor s neighbors whenever new data become available and the battery level is Portions of this chapter are adopted from two earlier books by the author 2 7 Fundamentals of Sensor Network Programming Applications and Technology By S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 41 www finebook ir 42 DATA STRUCTURES FOR SENSOR COMPUTING CPU DRIVEN EVENT DRIVEN EVENT STACK pp dme guns ESE pug c runt Busy wait b start for screen update event event available Li 1 1 1 Li H 1 H Li tS 1 H a M 1 Nou 1 N 1 a Busy wait for screen 1 1 1 1 1 Li ci Li 1 1 1 1 1 1 Li L k L 1 I a n Is Re charged update event ScreenSaver ScreenSaveri 0 Y Mq4 4 onldle
371. x reusability fusion classifiers for sensor data model Proc 2nd Int Conf Sensor Technologies and Applications SENSORCOMM 08 Aug 25 31 2008 pp 480 485 Iyer V S S Iyengar N Balakrishnan V Phoha and M B Srinivas FARMs Fusionable ambient renewable MACs Proc IEEE Sensors Applications Symp SAS 2009 Feb 17 19 2009 pp 169 174 Iyer V S S Iyengar G Rama Murthy M B Srinivas and B Hochet Multi hop scheduling and local datalink aggregation dependent qos in modeling and simulation of power aware wireless sensor networks Proceedings of 2009 ACM IWCMC Leipzig Germany 2009 pp 844 848 Iyer V G Rama Murthy M B Srinivas and B Hochet C error simulator for development for sensor and location aware sensing applications Proc 3rd Int Conf Sensing Technology ICST 2008 Nov 30 Dec 3 2008 pp 192 199 Iyer V R Murthy M B Srivinas and B Hochet Training data compression algorithms and reliability in large wireless sensor networks SUTC Proc IEEE Int Conf Sensor Networks Ubiquitous and Trustworthy Computing June 2008 pp 480 485 Iyer V G Rama Murthy and M B Srinivas Training data compression algorithms and relia bility in large wireless sensor networks Proc IEEE Int Conf Sensor Networks Ubiquitous and Trustworthy Computing June 2008 pp 480 485 Iyer V G Rama Murthy and M B Srinivas Environmental measurement OS for a tiny CRF stack used in wireless netwo
372. y computation storage and bandwidth constraints and 2 their overall usage scenario is quite different from that of traditional networks There is not a mere exchange of two specific nodes The user will be interested in some parameters of a dynamic physical process To efficiently achieve this the nodes have to form an application specific distributed system to provide the user with the answers More over simulation results show that the local computation takes a fraction of the energy when compared to data transmission So the design uses fixed resources such as a battery to do only sensing and local computations and uses a renewable energy resource to transmit data The nodes that are involved in the process of providing data information are constantly changing as the physical phenomenon is changing Therefore the user interacts with the system as a whole which provides sufficient information without loosing its usefulness drain due to power The WASN is not there to be queried but instead provides information in an efficient reliable and collaborative mode wirelessely Advancements in wireless technologies have made it possible to integrate with radios that are capable of self organizing and communicating in an ad hoc config uration with minimal infrastructure Our system INSPIRE maintains a real time network and periodically senses data from the onboard analog sensors which collec tively aggregate the sensed data by collaborative
373. y S S Iyengar N Parameshwaran V V Phoha N Balakrishnan and C D Okoye Copyright O 2011 John Wiley amp Sons Inc 155 www finebook ir 156 TECHNIQUES FOR PROTOCOL PROGRAMMING In the following sections we discuss the programming challenges involved in implementing a few basic protocols from the WSN 3 area 9 1 THE MEDIATION DEVICE PROTOCOL We now consider a well known protocol called the mediation device protocol In this protocol node A wants to transmit a packet to node B 1 Node A announces this to the mediation device by periodically sending a request to send RTS packets which the mediation device captures Node A sends its RTS packets instead of its query beacons and thus they have the same period 2 Node A then goes to receive mode listens 3 The mediation device waits for B s query and replies to B with a query response packet indicating A s address and a timing offset 4 Node B now knows when A will be listening again and sends a CTS to A Now B also knows A s period and thus knows at what time A s transmit mode will occur again 5 Node B waits at to receive data from A 6 At f B will send its acknowledgment packet B knows Z since it knows A s period 7 After the transaction has finished A restores its periodic wakeup cycle and starts to emit query beacons again 8 Node B also restores its own periodic cycle and thus decouples from A s period In the simple protocol shown
374. y features such as data confidentiality which includes encryption data with shared key data authentication and integrity This scheme allows the receiver to verify whether the data were sent by the client sender and also guarantees that messages are not altered in transit by hostile attackers One unique point of this algorithm is data freshness which guarantees that no adversaries replayed old sensor readings It also has a certain amount of odering of the sensor data due to data esimation A good mathematical theory has been developed for this sensor network encryption protocol for authenticated broadcast of network data For a broader treatment on this algorithm the reader can refer to the article by Perrig et al 5 16 7 FINAL COMMENTS Security in sensor networks is very critical to enhancing the long term usefulness of sensor networks for various applications This chapter has given a brief overview of the security aspects of these networks By no means is this a complete treatment of the subject matter Furthermore sensor networks is an area of national importance for many defense and civilian applications and cannot be considered deployable without sufficient protection from denial of service and other major attacks Consid eration of sensor network security in the design phase of the network can certainly ensure successful network deployment down the line and head off problems before they occur PROBLEMS 16 1 Develop a protoco
375. yer If it is a RTS frame the destination node checks whether its NAV timer is set whether its transmission region is busy and then responds to it by sending CTS All the other intermediate nodes receiving this RTS update their NAV timers to the CTS DATA ACK duration which implies that the channel is busy for that duration and hence they refrain from transmitting during this interval If the destination node receives more than two RTS requests within a given time interval then collision occurs and the destination node does not respond send CTS to any of these RTS requests The source nodes that are sending RTS have an RTSExpired timer set for RTS frames when they are sent to the destination node This timer is scheduled to expire after RTS CTS duration If the source node does not receive CTS within this duration RTSExpired timer expireds and the retry counter of that RTS frame is incremented If the retry counter is less than ShortRetieLimit as per the specification then the contention window is doubled and the random time set by the backoff timer is chosen between 1 and the contention window size If the retry counter reaches ShortRetryLimit then the message RTS and corresponding fragment is dropped by the MAC If the destination node responds to RTS by sending back the CTS the intermediate nodes for CTS will update their NAV timers obtained from the header field of the CTS frame DATA ACK duration and hence refrain from transmitting
376. yet it sends a negative ACK message to node A using the temporary address t Similarly other nodes also send negative ACK messages to A using t Node A waits for a sufficiently long time to receive these ACK messages and finally chooses f as its address Similarly node B generates a pair of random numbers lt t fi gt and floods the network Now node A will check whether f is the same as its address 1 if it is it sends a positive acknowl edgment otherwise it sends a negative acknowledgment Other nodes send positive acknowledgments If node A has sent a negative acknowledgment then node B on receiving this acknowledgment will generate a new pair of random numbers lt tj fi gt and flood the network Ultimately node B finds an address for ex ample 5 for itself This process continues until all nodes find unique addresses for themselves Of course one problem with this algorithm is that the nodes do not know how long each of them should wait for the acknowledgments Further the random number collisions can further delay the convergence of the algorithm The pseudocode for this algorithm is given below If we generate the addresses from large address spaces the collisions will be minimal We also assume that each node knows the total number of nodes present in the network assert f as own address Node P temp address pool T some addresses fixed address pool F ome addresses have address NO kmax 10 W
377. ying this pattern to real word scenarios is to maintain connectivity among nodes particularly when they are mobile With one or more fixed nodes and leader follower structures publish subscribe protocols and georouting geographic routing may be used When all nodes are mobile and if the structure is peer to peer structures such as Minimum Steiner Tree Approximated Minimum Steiner Tree or RoamHBA roaming hub based architec ture maintaining a roaming backbone may be used 10 4 PROGRAMMING BEYOND INDIVIDUAL NODES For large applications we need to combine the techniques mentioned above and write programs building abstractions on top of the patterns of collaboration discussed above This style of programming is not only scalable but also provides structural abstractions where each structure can be built designed implemented and maintained separately just as modular systems or object oriented systems PROBLEMS 10 1 List and explain some of the characteristics of wireless sensor networks 10 2 Describe data aggregation and the concept of tree data structure 10 3 Describe and list real life examples of publish subscribe systems and how the mechanism can be adapted to sensor networks www finebook ir 206 SENSOR DEPLOYMENT ABSTRACTION REFERENCES 1 V Iyer S S Iyengar N Balakrishnan V Phoha and M B Srinivas Farms Fusionable ambient renewable MACs Proc IEEE Sensors Applications Symp SAS 2009 Feb 17 19 2009

Fundamentals of Sensor Network Programming: Applications and

Contents

Download Pdf Manuals

Related Search

Related Contents